projekt_ttaras.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Przewidywanie dostępności stacjonarnego internetu szerokopasmowego na podstawie lokalizacji\n",
    "\n",
    "autor: Tomasz Taraś\n",
    "\n",
    "## Spis treści\n",
    "* Wstęp\n",
    "* Zbiory danych wykorzystane w pracy\n",
    "    * Dane o infrastrukturze szerokopasmowej i zasięgach (SIIS)\n",
    "    * Dane o adresach z Państwowego Rejestru Granic (PRG)\n",
    "    * Modele 3D budynków (LOD1)\n",
    "    * Numeryczny Model Terenu (NMT)\n",
    "    * Dane o nazwach urzędowych jednostek administracyjnych, miejscowości i ulic (rejestr TERYT)\n",
    "* Przygotowanie danych o infrastrukturze szerokopasmowej\n",
    "    * Zasięgi\n",
    "    * Węzły i interfejsy\n",
    "* Przygotowanie danych adresowych wraz z potrzebnymi dla modelu informacjami o charakterystyce przestrzennej danego punktu\n",
    "    * Parsowanie i czyszczenie danych adresowych z Państwowego Rejestru Granic (PRG)\n",
    "    * Przygotowanie obrysów budynków na podstawie modeli 3D (LOD1)\n",
    "    * Przygotowanie danych z Numerycznego Modelu Terenu\n",
    "    * Połączenie danych w spójny słownik adresów wraz z dodatkowymi informacjami\n",
    "* Przygotowanie kompletnego zbioru do uczenia modelu\n",
    "* Model uczenia maszynowego\n",
    "* Propozycje przyszłych działań\n",
    "* Podsumowanie\n",
    "\n",
    "## Wstęp\n",
    "\n",
    "Celem pracy jest przygotowanie modelu, który determinował będzie dostępność internetu na podstawie lokalizacji przestrzennej.\n",
    "\n",
    "Publicznie dostępne są dane na temat tego na jakie (prędkość, technologia) połączenie do internetu możemy liczyć w danej lokalizacji. Dane te zbierane są corocznie w pierwszym kwartale roku według stanu na 31 grudnia roku poprzedniego. Zbiera je Urząd Komunikacji Elektronicznej od przedsiębiorców telekomunikacyjnych. \n",
    "\n",
    "Dane te jednak zbierane są \"do adresu\" czyli jako lokalizacja jest adres budynku lub jego współrzędne, a rejestry adresowe w Polsce są średniej jakości z różnych względów. Przedsiębiorca nie ma możliwości podawania zasięgu swojej sieci w postaci np. poligonów określających przestrzennie skrawek terenu gdzie dany operator może świadczyć usługi. Dodatkowo rozporządzenie określające jakiego rodzaju dane przedsiębiorcy mają podawać w sprawozdaniu do Urzędu jest dość skomplikowane i budzi sporo wątpliwości wśród sprawozdających się.\n",
    "\n",
    "Te dwa czynniki powodują, że jakość danych nie jest idealna. Model stworzony w ramach tej pracy mogłby służyć jako imputer uzupełniający brakujące dane przy innych analizach lub do wykrywania anomalii, czyli obszarów, które dane wskazują jako pozbawione dostępu do szybkiego internetu, które według modelu powinny być w zasięgu sieci lub odwrotnie.\n",
    "\n",
    "Praca skupi się głównie na zebraniu i przygotowaniu danych co jest tutaj głównym wyzwaniem. Część danych, które mogłyby być przydatne do tego zadania jest niedostępna lub płatna i trzeba kreatywnie wykorzystać te zbiory, które są dostępne. Dodatkowym wyzwaniem, będzie rozmiar wykorzystywanych danych jako, że model do wytrenowania użyje danych z całej Polski, która ma ponad 7,4 mln adresów, ponad 14 mln budynków i ponad 311 tys. km2 powierzchni.\n",
    "\n",
    "Dostawcy internetu naturalnie najchętniej inwestują tam gdzie koszt jest niski. Im mniej infrastruktury trzeba dobudować dla klienta tym bardziej się to opłaca, dlatego miejsca o dużej gęstości zabudowy i budynki wielorodzinne mają dużo większą szansę na podłączenie światłowodu lub kablówki (lub ethernet), które zapewniają największe przepustowości. Dodatkowo dla technologii xDSL dostępna przepustowość jest zależna od odległości od węzła dlatego np Orange w jednym miejscu na kablu telefonicznym oferuje 10 mbit/s, a w innym również na kablu telefonicznym 80 mbit/s.\n",
    "\n",
    "Dataset, który zostanie przygotowany będzie zawierał informacje o zabudowie w postaci liczby sąsiadów oraz wielkości najbliższego budynku oraz odległość do najbliżsego węzła telekomunikacyjnego i jak dużo tych węzłów jest w okolicy. Dodatkowo będzie zawierał informacje o ukształtowaniu terenu wyrażonych jako wysokość nad poziomem morza i zmiana wysokości między najniższym punktem pomiarowym, a najwyższym w okolicy.\n",
    "\n",
    "## Zbiory danych wykorzystane w pracy\n",
    "\n",
    "W pracy wykorzystano następujące zbiory danych:\n",
    "\n",
    "#### Dane o infrastrukturze szerokopasmowej i zasięgach (SIIS)\n",
    "__podmiot__: Urząd Komunikacji Elektronicznej  \n",
    "__link__: [dane.gov.pl](https://dane.gov.pl/dataset/588,system-informacyjny-o-infrastrukturze-szerokopasmowej-api?page=1&per_page=50&sort=-title)  \n",
    "__rozmiar__: spakowane ok. 1GB, rozpakowane pliki CSV ok. 8GB  \n",
    "__opis__: Zbiór zawiera dane o infrastrukturze szerokopasmowej oraz o \"zasięgach\", czyli dostępności usługi dostępu do internetu. \n",
    "\n",
    "#### Dane o adresach z Państwowego Rejestru Granic (PRG)\n",
    "__podmiot__: Główny Urząd Geodezji i Kartografii  \n",
    "__link__: [gugik.gov.pl](http://www.gugik.gov.pl/pzgik/dane-bez-oplat/dane-z-panstwowego-rejestru-granic-i-powierzchni-jednostek-podzialow-terytorialnych-kraju-prg)  \n",
    "__rozmiar__: spakowane ok. 1GB, rozpakowane pliki XML ok. 18GB  \n",
    "__opis__: Zbiór zawiera listę adresów dla całęgo kraju zgodnie z tym co przekazane zostało do urzędu przez samorządy (gminy).\n",
    "\n",
    "#### Modele 3D budynków (LOD1)\n",
    "__podmiot__: Główny Urząd Geodezji i Kartografii  \n",
    "__link__: [linki w geoportalu krajowym](https://mapy.geoportal.gov.pl/imap/Imgp_2.html?locale=pl&gui=new&sessionID=4857706), [opis](https://integracja.gugik.gov.pl/Budynki3D/budynki3d_opis.pdf)  \n",
    "__rozmiar__: spakowane ok. 2,6 GB, rozpakowane pliki XML ok. 54 GB  \n",
    "__opis__: Zbiór zawiera modele 3D budynków o dokładności LOD1 dla całego kraju. Modele te zostały wykorzystane do utworzenia warstwy obrysów budynków 2D. To pozwoliło policzyć przybliżoną powierzchnię budynku dla adresu.  \n",
    "\n",
    "Informacje wymagane do podania zgodnie z licenją:  \n",
    "* Źródło: www.geoportal.gov.pl\n",
    "* Dysponent: Główny Geodeta Kraju\n",
    "* Data pobrania zbioru: 2019-11-10\n",
    "* Zakres przetworzenia: Geometria budynków została spłaszczona do 2D oraz wyekstrahowana została część poligonowa wykorzystana dalej jako obrys budynku.\n",
    "* Informacja: Modele 3D budynków nie stanowią rejestru publicznego ani elementu treści takiego rejestru. W konsekwencji czego mają wartość jedynie poglądową. Niezgodność Modeli 3D budynków ze stanem faktycznym lub prawnym, tak w postaci nieprzetworzonej jak i po ich ewentualnym przetworzeniu w procesie ponownego wykorzystania, nie może stanowić podstawy odpowiedzialności Głównego Geodety Kraju z jakiegokolwiek tytułu wobec jakiegokolwiek podmiotu.\n",
    "* Licencja: https://integracja.gugik.gov.pl/Budynki3D/GUGiK_Licencja_na_Budynki3D.pdf\n",
    "* [opis](https://integracja.gugik.gov.pl/Budynki3D/budynki3d_opis.pdf)\n",
    "\n",
    "#### Numeryczny Model Terenu (NMT)\n",
    "__podmiot__: Główny Urząd Geodezji i Kartografii  \n",
    "__link__: [gugik.gov.pl](http://www.gugik.gov.pl/pzgik/dane-bez-oplat/dane-dotyczace-numerycznego-modelu-terenu-o-interwale-siatki-co-najmniej-100-m-nmt_100)  \n",
    "__rozmiar__: spakowane ok. 122 MB, rozpakowane pliki txt ok. 826 MB  \n",
    "__opis__: Zbiór zawiera punkty pomiarowe (co 100m) wysokości terenu nad poziomem morza. Zostanie wykorzystany do określenia ukształtowania terenu wokół punktu.\n",
    "\n",
    "#### Dane o nazwach urzędowych jednostek administracyjnych, miejscowości i ulic (rejestr TERYT)\n",
    "__podmiot__: Główny Urząd Statystyczny  \n",
    "__link__: [eteryt.stat.gov.pl](http://eteryt.stat.gov.pl/eTeryt/rejestr_teryt/udostepnianie_danych/baza_teryt/uzytkownicy_indywidualni/pobieranie/pliki_pelne.aspx)  \n",
    "__rozmiar__: spakowane ok. 7MB, rozpakowane pliki CSV ok. 20MB  \n",
    "__opis__: Zbiór zawiera oficjalną listę nazw jednostek administracyjnych, miejscowości i ulic kompilowaną przez GUS na podstawie rozporządzeń (jednostki, miejscowości) oraz uchwał samorządowych (ulice). Zbiór wykorzystany do walidacj danych adresowych PRG.\n",
    "\n",
    "Wszystkie zbiory poza budynkami 3D należą do danych publicznych udostępnianych przez urzędy na podstawie odpowiednich przepisów. Budynki 3D udostępniane są na podstawie licencji podanej przy opisie zbioru.\n",
    "\n",
    "## Przygotowanie danych o infrastrukturze szerokopasmowej\n",
    "\n",
    "Po pobraniu archiwum z najnowszymi dostępnymi danymi ze strony: i rozpakowaniu go, ładujemy biblioteki i ścieżki do plików z danymi, żeby móc wstępnie przetowrzyć dane w Pythonie.\n",
    "\n",
    "Archiwum zawiera kilka plików CSV:\n",
    "\n",
    "* networkendpoint_view.csv - plik zawiera tzw. \"zakończenia sieci\" w praktyce oznacza to \"zasięg\" sieci w postaci listy adresów lub współrzędnych z informacją w jakiej technologii i o jakiej maksymalnej prędkości dostawca internetu może świadczyć usługę w danym budynku;\n",
    "\n",
    "* users_entity_view.csv - plik zawiera listę przedsiębiorców telekomunikacyjnych z id, służy głównie jako lookup table dla pozostałych plików, jeżeli interesuje nas który przedsiębiorca świadczy usługę lub posiada dany kawałek infrastruktury;\n",
    "\n",
    "* infrastructure_node_view.csv - plik zawiera informacje o węzłach telekomunikacyjnych, są to adresy lub współrzędne miejsc, gdzie jest aktywna infrastruktura telekomunikacyjna taka jak routery, switche itp;\n",
    "\n",
    "* infrastructure_nodeinterface_view.csv - plik zawiera informacje o interfejsach w węzłach, czyli sprzęt jakich technologii znajduje się w węźle;\n",
    "\n",
    "* infrastructure_noderangeradio_view.csv - plik zawiera informacje o zasięgach anten radiowych, na cele tej analizy nie będzie to nam potrzebne."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "folder = r'C:\\Users\\ttaras\\jupyter\\2018'\n",
    "path_nodes = os.path.join(folder, 'infrastructure_node_view.csv')\n",
    "path_node_interfaces = os.path.join(folder, 'infrastructure_nodeinterface_view.csv')\n",
    "path_node_range_radio = os.path.join(folder, 'infrastructure_noderangeradio_view.csv')\n",
    "path_companies = os.path.join(folder, 'users_entity_view.csv')\n",
    "path_endpoints = os.path.join(folder, 'networkendpoint_view.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Dalej przygotujemy oddzielnie dane dla zasięgów oraz dane dla węzłów.\n",
    "\n",
    "### Zasięgi\n",
    "\n",
    "Ładujemy dane o zasięgach z pliku CSV. Pomijamy od razu część kolumn i nadajemy odpowiednie typy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "endpoints_df = pd.read_csv(\n",
    "    path_endpoints, \n",
    "    usecols=[\n",
    "        'municipality_code_value',\n",
    "        'city_code_value',\n",
    "        'city_name',\n",
    "        'street_code_value',\n",
    "        'street_name',\n",
    "        'house_no',\n",
    "        'latitude',\n",
    "        'longitude',\n",
    "        'medium',\n",
    "        'stationary_bandwidth_name'\n",
    "    ],\n",
    "    dtype={\n",
    "        'municipality_code_value': 'category',\n",
    "        'city_code_value': 'category',\n",
    "        'city_name': str,\n",
    "        'street_code_value': 'category',\n",
    "        'street_name': str,\n",
    "        'house_no': str,\n",
    "        'latitude': np.float64,\n",
    "        'longitude': np.float64,\n",
    "        'medium': 'category',\n",
    "        'stationary_bandwidth_name': np.float32\n",
    "    }\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tak wygląda DataFrame po załadowaniu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>municipality_code_value</th>\n",
       "      <th>city_code_value</th>\n",
       "      <th>city_name</th>\n",
       "      <th>street_code_value</th>\n",
       "      <th>street_name</th>\n",
       "      <th>house_no</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "      <th>medium</th>\n",
       "      <th>stationary_bandwidth_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1418044</td>\n",
       "      <td>0921438</td>\n",
       "      <td>Piaseczno</td>\n",
       "      <td>09340</td>\n",
       "      <td>ul. Kordeckiego</td>\n",
       "      <td>4</td>\n",
       "      <td>52.057000</td>\n",
       "      <td>21.002062</td>\n",
       "      <td>radiowe</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1418044</td>\n",
       "      <td>0921438</td>\n",
       "      <td>Piaseczno</td>\n",
       "      <td>10036</td>\n",
       "      <td>ul. Krupówki</td>\n",
       "      <td>7</td>\n",
       "      <td>52.068033</td>\n",
       "      <td>21.033234</td>\n",
       "      <td>radiowe</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1418044</td>\n",
       "      <td>0921438</td>\n",
       "      <td>Piaseczno</td>\n",
       "      <td>11205</td>\n",
       "      <td>ul. 11 Listopada</td>\n",
       "      <td>83</td>\n",
       "      <td>52.050284</td>\n",
       "      <td>20.990206</td>\n",
       "      <td>radiowe</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1418044</td>\n",
       "      <td>0921438</td>\n",
       "      <td>Piaseczno</td>\n",
       "      <td>14907</td>\n",
       "      <td>ul. Okrężna</td>\n",
       "      <td>18</td>\n",
       "      <td>52.062186</td>\n",
       "      <td>20.982096</td>\n",
       "      <td>radiowe</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1418044</td>\n",
       "      <td>0921438</td>\n",
       "      <td>Piaseczno</td>\n",
       "      <td>15129</td>\n",
       "      <td>ul. Orężna</td>\n",
       "      <td>7</td>\n",
       "      <td>52.070958</td>\n",
       "      <td>21.007444</td>\n",
       "      <td>radiowe</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  municipality_code_value city_code_value  city_name street_code_value  \\\n",
       "0                 1418044         0921438  Piaseczno             09340   \n",
       "1                 1418044         0921438  Piaseczno             10036   \n",
       "2                 1418044         0921438  Piaseczno             11205   \n",
       "3                 1418044         0921438  Piaseczno             14907   \n",
       "4                 1418044         0921438  Piaseczno             15129   \n",
       "\n",
       "        street_name house_no   latitude  longitude   medium  \\\n",
       "0   ul. Kordeckiego        4  52.057000  21.002062  radiowe   \n",
       "1      ul. Krupówki        7  52.068033  21.033234  radiowe   \n",
       "2  ul. 11 Listopada       83  52.050284  20.990206  radiowe   \n",
       "3       ul. Okrężna       18  52.062186  20.982096  radiowe   \n",
       "4        ul. Orężna        7  52.070958  21.007444  radiowe   \n",
       "\n",
       "   stationary_bandwidth_name  \n",
       "0                        0.0  \n",
       "1                        0.0  \n",
       "2                        0.0  \n",
       "3                        0.0  \n",
       "4                        0.0  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "endpoints_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plik zawiera informacje o zasięgach stacjonarnych oraz mobilnych (LTE etc.). Zostawimy sobie tylko interesujące nas zasięgi stacjonarne."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ogólna liczba wierszy w pliku CSV: 44123969\n",
      "Liczba wierszy po odfiltrowaniu zasięgów mobilnych: 11011310\n"
     ]
    }
   ],
   "source": [
    "total_row_count = len(endpoints_df.index)\n",
    "endpoints_df = endpoints_df[endpoints_df['stationary_bandwidth_name'] > 0]\n",
    "filtered_row_count = len(endpoints_df.index)\n",
    "\n",
    "print('Ogólna liczba wierszy w pliku CSV:', total_row_count)\n",
    "print('Liczba wierszy po odfiltrowaniu zasięgów mobilnych:', filtered_row_count)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Większość rekordów jest zestandaryzowana i zawiera kody gmin, miejscowości i ulic zgodne z urzędowym rejestrem TERYT, ale część np. ulic z różnych powodów nie ma tych kodów podanych, a zamiast tego jest kod stanowiący placeholder: '99998'. Sprawdźmy najpierw ile takich rekordów jest w naszych danych:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "17582"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(endpoints_df[endpoints_df['street_code_value'] == '99998'].index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Niewiele, ale prawdopodobnie niewielkim kosztem będziemy w stanie część z nich uzupełnić, więc spróbujemy.\n",
    "\n",
    "Innym kodem stanowiącym placeholder jest kod ulic: '99999', który oznacza brak nazwy ulicy. W miastach się to nie zdarza, ale mniejsze miejscowości i wsie dość często nie nadają nazw ulicom/drogom i adres to po prostu miejscowość i numer porządkowy np. Stara Wieś 3.\n",
    "\n",
    "Przeprowadzimy teraz drobne czyszczenie i standaryzację danych.\n",
    "\n",
    "Najpierw przeprowadzimy operację trim/strip usuwającą zbędne znaki białe z początku i końca wartości tekstowych, a następnie wszystkie litery występujące w numerach porządkowych zamienimy na wielkie, a nazwy ulic na małe. Dodatkowo z nazw ulic usuniemy przedrostek 'ul. '. Te operacje pozwolą nam porównać ten zbiór danych z innymi."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "endpoints_df['house_no'] = endpoints_df['house_no'].str.strip().str.upper()\n",
    "endpoints_df['city_name'] = endpoints_df['city_name'].str.strip().str.lower()\n",
    "endpoints_df['street_name'] = endpoints_df['street_name'].str.strip().str.lower()\n",
    "endpoints_df['street_name'] = endpoints_df['street_name'].str.replace('ul. ', '')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Następnie dla zasięgów radiowych wszystkie wartości ponad 50 mbit/s sprowadzimy do wartości 50 mbit/s. Technicznie mało prawdopodobne, żeby dostęp radiowy pozwalał na szybsze prędkości o ile nie jest to LTE ze świetnym zasięgiem lub dedykowana radiolinia i większość z tych przypadków to najprawdopoboniej błędy w danych lub błędne interpretacje jakie dane powinny być wysłane do urzędu w ramach sprawozdania."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "endpoints_df.loc[\n",
    "    (endpoints_df['medium'] == 'radiowe') & (endpoints_df['stationary_bandwidth_name'] > 50), \n",
    "    'stationary_bandwidth_name'\n",
    "] = 50"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Teraz spróbujemy uzupełnić część brakujących kodów ulic posiłkujac się istniejącymi w innych rekordach nazwami ulic i ich kodami. Następnie odrzucamy te rekordy których nie udało nam się dopasować."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "lookup = endpoints_df[~endpoints_df['street_code_value'].isin({'99999', '99998'})][['city_code_value', 'street_code_value', 'street_name']].drop_duplicates().dropna()\n",
    "\n",
    "lookup.set_index(['city_code_value', 'street_name'], inplace=True)\n",
    "\n",
    "temp = endpoints_df[endpoints_df['street_code_value'] == '99998'][['city_code_value', 'street_name']].join(lookup, on=['city_code_value', 'street_name'], rsuffix='_joined')\n",
    "\n",
    "endpoints_df = endpoints_df.join(temp, rsuffix='_joined')\n",
    "\n",
    "endpoints_df['street_code_value'] = endpoints_df[['street_code_value', 'street_code_value_joined']].apply(lambda x: x[0] if x[0] != '99998' else x[1], axis=1)\n",
    "\n",
    "endpoints_df.drop(\n",
    "    columns=['city_code_value_joined', 'street_code_value_joined', 'street_name_joined'], \n",
    "    inplace=True\n",
    ")\n",
    "\n",
    "endpoints_df.dropna(\n",
    "    subset=['street_code_value', 'street_name'], \n",
    "    how='all', \n",
    "    axis=0, \n",
    "    inplace=True\n",
    ")\n",
    "\n",
    "endpoints_df = endpoints_df[endpoints_df['street_code_value'] != '99998']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sprawdzamy ile rekordów nam zostało:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "10947485"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(endpoints_df.index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Usuwamy współrzędne z rekordów gdzie są podane liczby całkowite. Astronomicznie mało prawdopodobne żeby współrzędne jakiegoś obiektu to było dokładnie 20, 50 zamiast 20.000034, 50.000011. Prawie na pewno są to błędy w sprawozdawczości przedsiębiorców.\n",
    "\n",
    "Współrzędne tutaj będziemy traktować drugorzędnie. Jako podstawowe źródło posłuży nam zbiór PRG z urzędowymi adresami."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "endpoints_df.loc[(endpoints_df['latitude'] == round(endpoints_df['latitude'])) & (endpoints_df['longitude'] == round(endpoints_df['longitude'])), ['latitude', 'longitude']] = np.nan"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tak wygląda nasz DataFrame przed eksportem:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>municipality_code_value</th>\n",
       "      <th>city_code_value</th>\n",
       "      <th>city_name</th>\n",
       "      <th>street_code_value</th>\n",
       "      <th>street_name</th>\n",
       "      <th>house_no</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "      <th>medium</th>\n",
       "      <th>stationary_bandwidth_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>626</th>\n",
       "      <td>1432064</td>\n",
       "      <td>0921415</td>\n",
       "      <td>ożarów mazowiecki</td>\n",
       "      <td>17394</td>\n",
       "      <td>poznańska</td>\n",
       "      <td>167</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>kablowe współosiowe miedziane</td>\n",
       "      <td>250.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>627</th>\n",
       "      <td>1432064</td>\n",
       "      <td>0921415</td>\n",
       "      <td>ożarów mazowiecki</td>\n",
       "      <td>12740</td>\n",
       "      <td>adama mickiewicza</td>\n",
       "      <td>7</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>światłowodowe</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>628</th>\n",
       "      <td>1432064</td>\n",
       "      <td>0921415</td>\n",
       "      <td>ożarów mazowiecki</td>\n",
       "      <td>12740</td>\n",
       "      <td>adama mickiewicza</td>\n",
       "      <td>5A</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>światłowodowe</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1259</th>\n",
       "      <td>1201011</td>\n",
       "      <td>0981682</td>\n",
       "      <td>bochnia</td>\n",
       "      <td>24983</td>\n",
       "      <td>wygoda</td>\n",
       "      <td>64</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>radiowe</td>\n",
       "      <td>50.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1385</th>\n",
       "      <td>0603011</td>\n",
       "      <td>0930176</td>\n",
       "      <td>rejowiec fabryczny</td>\n",
       "      <td>10562</td>\n",
       "      <td>kwiatowa</td>\n",
       "      <td>72</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>światłowodowe</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     municipality_code_value city_code_value           city_name  \\\n",
       "626                  1432064         0921415   ożarów mazowiecki   \n",
       "627                  1432064         0921415   ożarów mazowiecki   \n",
       "628                  1432064         0921415   ożarów mazowiecki   \n",
       "1259                 1201011         0981682             bochnia   \n",
       "1385                 0603011         0930176  rejowiec fabryczny   \n",
       "\n",
       "     street_code_value        street_name house_no  latitude  longitude  \\\n",
       "626              17394          poznańska      167       NaN        NaN   \n",
       "627              12740  adama mickiewicza        7       NaN        NaN   \n",
       "628              12740  adama mickiewicza       5A       NaN        NaN   \n",
       "1259             24983             wygoda       64       NaN        NaN   \n",
       "1385             10562           kwiatowa       72       NaN        NaN   \n",
       "\n",
       "                             medium  stationary_bandwidth_name  \n",
       "626   kablowe współosiowe miedziane                      250.0  \n",
       "627                   światłowodowe                     1000.0  \n",
       "628                   światłowodowe                     1000.0  \n",
       "1259                        radiowe                       50.0  \n",
       "1385                  światłowodowe                     1000.0  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "endpoints_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Eksport danych o zasięgach do pliku CSV."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "endpoints_df.to_csv(r'C:\\Users\\ttaras\\jupyter\\zasiegi.csv', sep=';', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Węzły i interfejsy\n",
    "\n",
    "Ładujemy dane o węzłach i interfejsach z plików CSV. Pomijamy od razu część kolumn i nadajemy odpowiednie typy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "nodes_df = pd.read_csv(\n",
    "    path_nodes, \n",
    "    usecols=[\n",
    "        'municipality_code_value',\n",
    "        'city_code_value',\n",
    "        'city_name',\n",
    "        'street_code_value',\n",
    "        'street_name',\n",
    "        'house_no',\n",
    "        'latitude',\n",
    "        'longitude',\n",
    "        'id'\n",
    "    ],\n",
    "    dtype={\n",
    "        'municipality_code_value': str,\n",
    "        'city_code_value': str,\n",
    "        'city_name': str,\n",
    "        'street_code_value': str,\n",
    "        'street_name': str,\n",
    "        'house_no': str,\n",
    "        'latitude': np.float64,\n",
    "        'longitude': np.float64,\n",
    "    }\n",
    ")\n",
    "\n",
    "nodes_df.rename(columns={'id': 'node_id'}, inplace=True)\n",
    "nodes_df.set_index('node_id', inplace=True)\n",
    "\n",
    "int_df = pd.read_csv(\n",
    "    path_node_interfaces, \n",
    "    usecols=[\n",
    "        'node_id', \n",
    "        'transmission_medium', \n",
    "        'technology_twisted_pair_copper'\n",
    "    ]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tak wyglądają nasze DataFrame po załadowaniu:\n",
    "\n",
    "Węzły:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>municipality_code_value</th>\n",
       "      <th>city_code_value</th>\n",
       "      <th>city_name</th>\n",
       "      <th>street_code_value</th>\n",
       "      <th>street_name</th>\n",
       "      <th>house_no</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>node_id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3355763</th>\n",
       "      <td>3001022</td>\n",
       "      <td>0524542</td>\n",
       "      <td>Wyszyny</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>52.8923</td>\n",
       "      <td>16.8586</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5233033</th>\n",
       "      <td>0662011</td>\n",
       "      <td>0929902</td>\n",
       "      <td>Chełm</td>\n",
       "      <td>21787</td>\n",
       "      <td>ul. Szarych Szeregów</td>\n",
       "      <td>9B</td>\n",
       "      <td>51.1421</td>\n",
       "      <td>23.4300</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5233032</th>\n",
       "      <td>1425062</td>\n",
       "      <td>0625668</td>\n",
       "      <td>Słupica</td>\n",
       "      <td>99999</td>\n",
       "      <td>NaN</td>\n",
       "      <td>168</td>\n",
       "      <td>51.4124</td>\n",
       "      <td>21.4048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5233029</th>\n",
       "      <td>0662011</td>\n",
       "      <td>0929902</td>\n",
       "      <td>Chełm</td>\n",
       "      <td>11205</td>\n",
       "      <td>ul. 11 Listopada</td>\n",
       "      <td>2</td>\n",
       "      <td>51.8325</td>\n",
       "      <td>23.2944</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5233030</th>\n",
       "      <td>0603114</td>\n",
       "      <td>0107821</td>\n",
       "      <td>Siedliszcze</td>\n",
       "      <td>22073</td>\n",
       "      <td>ul. Szpitalna</td>\n",
       "      <td>15A</td>\n",
       "      <td>51.1963</td>\n",
       "      <td>23.1603</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        municipality_code_value city_code_value    city_name  \\\n",
       "node_id                                                        \n",
       "3355763                 3001022         0524542      Wyszyny   \n",
       "5233033                 0662011         0929902        Chełm   \n",
       "5233032                 1425062         0625668      Słupica   \n",
       "5233029                 0662011         0929902        Chełm   \n",
       "5233030                 0603114         0107821  Siedliszcze   \n",
       "\n",
       "        street_code_value           street_name house_no  latitude  longitude  \n",
       "node_id                                                                        \n",
       "3355763               NaN                   NaN      NaN   52.8923    16.8586  \n",
       "5233033             21787  ul. Szarych Szeregów       9B   51.1421    23.4300  \n",
       "5233032             99999                   NaN      168   51.4124    21.4048  \n",
       "5233029             11205      ul. 11 Listopada        2   51.8325    23.2944  \n",
       "5233030             22073         ul. Szpitalna      15A   51.1963    23.1603  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nodes_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Interfejsy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>node_id</th>\n",
       "      <th>transmission_medium</th>\n",
       "      <th>technology_twisted_pair_copper</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>15837612</td>\n",
       "      <td>światłowodowe</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>12144737</td>\n",
       "      <td>kablowe parowe miedziane</td>\n",
       "      <td>1 Gigabit Ethernet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>14365015</td>\n",
       "      <td>kablowe parowe miedziane</td>\n",
       "      <td>1 Gigabit Ethernet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>14822051</td>\n",
       "      <td>radiowe</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2917895</td>\n",
       "      <td>światłowodowe</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    node_id       transmission_medium technology_twisted_pair_copper\n",
       "0  15837612             światłowodowe                            NaN\n",
       "1  12144737  kablowe parowe miedziane             1 Gigabit Ethernet\n",
       "2  14365015  kablowe parowe miedziane             1 Gigabit Ethernet\n",
       "3  14822051                   radiowe                            NaN\n",
       "4   2917895             światłowodowe                            NaN"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "int_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Podobnie jak w przypadku zasięgów zestandaryzujemy adresy węzłów."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "nodes_df['house_no'] = nodes_df['house_no'].str.strip().str.upper()\n",
    "nodes_df['city_name'] = nodes_df['city_name'].str.strip().str.lower()\n",
    "nodes_df['street_name'] = nodes_df['street_name'].str.strip().str.lower()\n",
    "nodes_df['street_name'] = nodes_df['street_name'].str.replace('ul. ', '')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Następnie dla interfejsów rozpiszemy wartości z kolumny 'medium transymisyjne' (oraz 'technology_twisted_pair_copper') metodą one-hot encoding, czyli stworzymy 4 kolumny, które będą przybierać wartośc 0 lub 1 zależnie od tego czy dane medium np. światłowodowe odpowiada temu rekordowi czy nie. Pogrupujemy interfejsy po id węzła dzięki czemu uzyskamy informacje per węzeł na temat technologii jakie są dla niego dostępne.\n",
    "Dla medium kablowego parowego robimy wyjątek i dzielimy je na kable telefoniczne i nie-telefoniczne co odpowiada odpwiednio 1 i 0 w kolumnie 'xDSL'. Dla prostszego przetwarzania zostawiona zostanie kolumna 'copper' dla wszystkich kabli miedzianych."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "int_df['fiber'] = int_df['transmission_medium'].apply(lambda x: 1 if x == 'światłowodowe' else 0)\n",
    "\n",
    "int_df['xdsl'] = int_df['technology_twisted_pair_copper'].apply(lambda x: 1 if x in ('ADSL2+', 'POTS/ISDN', 'VDSL2', 'ADSL2', 'ADSL', 'VDSL') else 0)\n",
    "\n",
    "int_df['copper'] = int_df['transmission_medium'].apply(lambda x: 1 if x in ('kablowe parowe miedziane', 'kablowe współosiowe miedziane') else 0)\n",
    "\n",
    "int_df['radio'] = int_df['transmission_medium'].apply(lambda x: 1 if x == 'radiowe' else 0)\n",
    "\n",
    "int_df.drop(\n",
    "    columns=['transmission_medium', 'technology_twisted_pair_copper'], \n",
    "    inplace=True\n",
    ")\n",
    "\n",
    "int_df = int_df.groupby(by='node_id', as_index=True)[['fiber', 'copper', 'xdsl', 'radio']].max()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Dla węzłów też spróbujemy uzupełnić brakujące kody ulic."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "temp = nodes_df[nodes_df['street_code_value'] == '99998'][['city_code_value', 'street_name']].join(lookup, on=['city_code_value', 'street_name'], rsuffix='_joined')\n",
    "\n",
    "nodes_df = nodes_df.join(temp, rsuffix='_joined')\n",
    "\n",
    "nodes_df['street_code_value'] = nodes_df[['street_code_value', 'street_code_value_joined']].apply(lambda x: x[0] if x[0] != '99998' else x[1], axis=1)\n",
    "\n",
    "nodes_df.drop(\n",
    "    columns=['city_code_value_joined', 'street_code_value_joined', 'street_name_joined'], \n",
    "    inplace=True\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Usuwamy współrzędne tam gdzie są całkowite."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "nodes_df.loc[(nodes_df['latitude'] == round(nodes_df['latitude'])) & (nodes_df['longitude'] == round(nodes_df['longitude'])), ['latitude', 'longitude']] = np.nan"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Dołączamy do węzłów informacje o technologiach z interfejsów."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "nodes_df = nodes_df.join(int_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ile węzłów mamy w naszym DataFrame:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "348670"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(nodes_df.index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ostatecznie uzupełniamy jeszcze brakujące wartości dla kolumn oznaczających technologie. Brak wartości jest równoznaczny 0 zgodnie z naszym kodowaniem."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "nodes_df['fiber'].fillna(0, inplace=True)\n",
    "nodes_df['copper'].fillna(0, inplace=True)\n",
    "nodes_df['xdsl'].fillna(0, inplace=True)\n",
    "nodes_df['radio'].fillna(0, inplace=True)\n",
    "\n",
    "nodes_df = nodes_df.astype({\n",
    "    'fiber': 'int8',\n",
    "    'copper': 'int8',\n",
    "    'xdsl': 'int8',\n",
    "    'radio': 'int8'\n",
    "})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tak wygląda nasz DataFrame przed eksportem:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>municipality_code_value</th>\n",
       "      <th>city_code_value</th>\n",
       "      <th>city_name</th>\n",
       "      <th>street_code_value</th>\n",
       "      <th>street_name</th>\n",
       "      <th>house_no</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "      <th>fiber</th>\n",
       "      <th>copper</th>\n",
       "      <th>xdsl</th>\n",
       "      <th>radio</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>node_id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>86</th>\n",
       "      <td>1219055</td>\n",
       "      <td>0340180</td>\n",
       "      <td>czarnochowice</td>\n",
       "      <td>99999</td>\n",
       "      <td>NaN</td>\n",
       "      <td>146</td>\n",
       "      <td>50.0117</td>\n",
       "      <td>20.0728</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>87</th>\n",
       "      <td>1206172</td>\n",
       "      <td>0344136</td>\n",
       "      <td>brzozówka</td>\n",
       "      <td>20254</td>\n",
       "      <td>słoneczna</td>\n",
       "      <td>1</td>\n",
       "      <td>50.1771</td>\n",
       "      <td>19.8934</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>91</th>\n",
       "      <td>1219055</td>\n",
       "      <td>0340256</td>\n",
       "      <td>chorągwica</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>49.9581</td>\n",
       "      <td>20.0853</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>92</th>\n",
       "      <td>1219022</td>\n",
       "      <td>0318972</td>\n",
       "      <td>zręczyce</td>\n",
       "      <td>99999</td>\n",
       "      <td>NaN</td>\n",
       "      <td>91</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>94</th>\n",
       "      <td>1219012</td>\n",
       "      <td>0315175</td>\n",
       "      <td>przebieczany</td>\n",
       "      <td>99999</td>\n",
       "      <td>NaN</td>\n",
       "      <td>136</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        municipality_code_value city_code_value      city_name  \\\n",
       "node_id                                                          \n",
       "86                      1219055         0340180  czarnochowice   \n",
       "87                      1206172         0344136      brzozówka   \n",
       "91                      1219055         0340256     chorągwica   \n",
       "92                      1219022         0318972       zręczyce   \n",
       "94                      1219012         0315175   przebieczany   \n",
       "\n",
       "        street_code_value street_name house_no  latitude  longitude  fiber  \\\n",
       "node_id                                                                      \n",
       "86                  99999         NaN      146   50.0117    20.0728      0   \n",
       "87                  20254   słoneczna        1   50.1771    19.8934      0   \n",
       "91                    NaN         NaN      NaN   49.9581    20.0853      0   \n",
       "92                  99999         NaN       91       NaN        NaN      0   \n",
       "94                  99999         NaN      136       NaN        NaN      0   \n",
       "\n",
       "         copper  xdsl  radio  \n",
       "node_id                       \n",
       "86            0     0      1  \n",
       "87            0     0      1  \n",
       "91            0     0      1  \n",
       "92            0     0      1  \n",
       "94            0     0      1  "
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nodes_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Eksport danych o węzłach do pliku CSV."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "nodes_df.to_csv(r'C:\\Users\\ttaras\\jupyter\\wezly.csv', sep=';', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Przygotowanie danych adresowych wraz z potrzebnymi dla modelu informacjami o charakterystyce przestrzennej danego punktu\n",
    "\n",
    "### Parsowanie i czyszczenie danych adresowych z Państwowego Rejestru Granic (PRG)\n",
    "\n",
    "Dane udostępniane w ramach Państwowego Rejestru Granic to granice jednostek administracyjnych: kraj, województwa, powiaty, gminy oraz punkty adresowe. Dla tej pracy interesują nas punkty adresowe, które będą stanowić uzupełnienie danych dla zbioru SIIS o zasięgach sieci, a konkretnie pozwolą nam dodać brakujące współrzędne geograficzne. \n",
    "\n",
    "Dobrze by było wykorzystać zbiór adresów do dodania także adresów, które nie są w zasięgu żadnej sieci stacjonarnej, czyli nie będzie ich w zbiorze SIIS, ale trzeba by mieć w miarę zsynchronizowane czasowo zbiory, a najnowsza porcja zbioru SIIS zawiera dane za 2018 rok. W związku z tym nie będziemy dodawać adresów bez sieci, bo przez prawie dwuletnią różnicę sporo nowych adresów, które w rzeczywistości są w zasięgu sieci było by pokazane jako bez internetu co bardziej zaburzyło by wyniki modelu niż brak adresów bez internetu.\n",
    "\n",
    "PRG udostępniane jest w postaci spakowanych plików GML (rozszerzenie XML dla danych przestrzennych). Struktura wygląda tak, że w ramach pliku mamu kilka rodzajów rekordów: jednostkę administracyjną (dla plików adresowych używanych przez nas bez geometrii tylko dane tekstowe jak nazwa), miejscowość (nazwa, kod teryt), ulicę (nazwa, kod teryt, geometria) oraz adres, który zawiera informacje tekstowe o adresie, geometrię (współrzędne geograficzne), ale także łączy się z pozostałymi rodzajami rekordów. Czyli w ramach hierarhicznego typu plików jakim jest XML mamy zbiór relacyjny. Co ciekawe pomimo duplikacji części danych jak np. nazwa miejscowości czy ulicy dane te nie zawsze zgadzają się pomiędzy rekordem adresu, a odpowiadającym mu rekordem miejscowości/ulicy. Na podstawie wstępnej analizy danych doszedłem do wniosku, że lepiej będzie korzystać z nazw miejscowości/ulic pochodzących z rekordów tych typów połączonych z rekordem adresu niż z samych danych zawartych w rekordzie adresu.\n",
    "\n",
    "Przykładowy fragment xml:\n",
    "```xml\n",
    "<prg-ad:PRG_PunktAdresowy gml:id=\"PL.ZIPIN.3122.EMUiA_acfe21a9-1a6d-4143-b57f-6e96af120cd1\">\n",
    "      <gml:identifier codeSpace=\"http://geoportal.gov.pl/\">http://geoportal.gov.pl/PZGIK/dane/PL.ZIPIN.3122.EMUiA/acfe21a9-1a6d-4143-b57f-6e96af120cd1</gml:identifier>\n",
    "      <prg-ad:idIIP>\n",
    "        <bt:BT_Identyfikator>\n",
    "          <bt:lokalnyId>dc558f5b-bc07-47ea-bfb3-65b8314cfd54</bt:lokalnyId>\n",
    "          <bt:przestrzenNazw>PL.PZGIK.200</bt:przestrzenNazw>\n",
    "          <bt:wersjaId>2019-07-03T17:59:05+02:00</bt:wersjaId>\n",
    "        </bt:BT_Identyfikator>\n",
    "      </prg-ad:idIIP>\n",
    "      <prg-ad:cyklZycia>\n",
    "        <bt:BT_CyklZyciaInfo>\n",
    "          <bt:poczatekWersjiObiektu>2019-07-03T17:59:05Z</bt:poczatekWersjiObiektu>\n",
    "        </bt:BT_CyklZyciaInfo>\n",
    "      </prg-ad:cyklZycia>\n",
    "      <prg-ad:waznyOd>2019-07-03</prg-ad:waznyOd>\n",
    "      <prg-ad:jednostkaAdmnistracyjna>Polska</prg-ad:jednostkaAdmnistracyjna>\n",
    "      <prg-ad:jednostkaAdmnistracyjna>lubuskie</prg-ad:jednostkaAdmnistracyjna>\n",
    "      <prg-ad:jednostkaAdmnistracyjna>sulęciński</prg-ad:jednostkaAdmnistracyjna>\n",
    "      <prg-ad:jednostkaAdmnistracyjna>Lubniewice</prg-ad:jednostkaAdmnistracyjna>\n",
    "      <prg-ad:miejscowosc>Lubniewice</prg-ad:miejscowosc>\n",
    "      <prg-ad:czescMiejscowosci nilReason=\"inapplicable\" xsi:nil=\"true\" />\n",
    "      <prg-ad:ulica>Spacerowa</prg-ad:ulica>\n",
    "      <prg-ad:numerPorzadkowy>20</prg-ad:numerPorzadkowy>\n",
    "      <prg-ad:kodPocztowy>69-210</prg-ad:kodPocztowy>\n",
    "      <prg-ad:status>istniejacy</prg-ad:status>\n",
    "      <prg-ad:pozycja>\n",
    "        <gml:Point srsName=\"urn:ogc:def:crs:EPSG::2180\" srsDimension=\"2\" gml:id=\"PL.PZGIK.200_dc558f5b-bc07-47ea-bfb3-65b8314cfd54_GEO_1\">\n",
    "          <gml:pos>523018.0559 244304.4583</gml:pos>\n",
    "        </gml:Point>\n",
    "      </prg-ad:pozycja>\n",
    "      <prg-ad:komponent xlink:href=\"http://geoportal.gov.pl/PZGIK/dane/PL.PZGIK.200_366263\" />\n",
    "      <prg-ad:komponent xlink:href=\"http://geoportal.gov.pl/PZGIK/dane/PL.PZGIK.200_366267\" />\n",
    "      <prg-ad:komponent xlink:href=\"http://geoportal.gov.pl/PZGIK/dane/PL.PZGIK.200_366442\" />\n",
    "      <prg-ad:komponent xlink:href=\"http://geoportal.gov.pl/PZGIK/dane/PL.PZGIK.200_370131\" />\n",
    "      <prg-ad:komponent xlink:href=\"http://geoportal.gov.pl/PZGIK/dane/PL.ZIPIN.3122.EMUiA_0182969\" />\n",
    "      <prg-ad:komponent xlink:href=\"http://geoportal.gov.pl/PZGIK/dane/PL.ZIPIN.3122.EMUiA_3aadbcd4-aadc-47a0-98d0-057f9b2adde1\" />\n",
    "      <prg-ad:obiektEMUiA xlink:href=\"http://geoportal.gov.pl/PZGIK/dane/PL.ZIPIN.3122.EMUiA/acfe21a9-1a6d-4143-b57f-6e96af120cd1\" />\n",
    "    </prg-ad:PRG_PunktAdresowy>\n",
    "```\n",
    "\n",
    "W związku z rozmiarem plików nie możemy załadować ich po prostu do pamięci jak zazwyczaj robi się z plikami XML (parsowanie [DOM](https://en.wikipedia.org/wiki/Document_Object_Model)). Zastosujemy parsowanie \"zdarzeniowe\" ([SAX](https://en.wikipedia.org/wiki/Simple_API_for_XML)), dzięki któremu będziemy przetwarzać plik rekord po rekordzie. Dzięki temu zużycie pamięci będzie minimalne. Każdy rekord po przetworzeniu ładowany jest do bazy PostgreSQL, gdzie dokonamy dalszych przetworzeń i czyszczenia zbioru.\n",
    "\n",
    "Skrypt w python parsujący pliki PRG załączony jest w repozytorium git podobnie jak skrypt pobierający słowniki TERYT oraz kwerendy SQL, którymi dalej przetwarzano zbiór przygotowując go do połączenia z danymi SIIS.\n",
    "\n",
    "### Przygotowanie obrysów budynków na podstawie modeli 3D (LOD1)\n",
    "\n",
    "Dane są w formacie CityGML, który jest rozszerzeniem XML (GML). Ze względu na skomplikowaną strukturę użyto programu Spatial ETL o nazwie FME od firmy Safe Software.\n",
    "\n",
    "Przykładowy fragment xml:\n",
    "```xml\n",
    "<core:cityObjectMember>\n",
    "\t\t<bldg:Building gml:id=\"ID-1432-38E036F4-D4A5-9AC7-E053-CC2BA8C036B5\">\n",
    "\t\t\t<gen:stringAttribute name=\"buildingId\">\n",
    "\t\t\t\t<gen:value>38E036F4-D4A5-9AC7-E053-CC2BA8C036B5</gen:value>\n",
    "\t\t\t</gen:stringAttribute>\n",
    "\t\t\t<gen:stringAttribute name=\"przestNazw\">\n",
    "\t\t\t\t<gen:value>PL.PZGiK.330.BDOT10k</gen:value>\n",
    "\t\t\t</gen:stringAttribute>\n",
    "\t\t\t<gen:stringAttribute name=\"wersjaId\">\n",
    "\t\t\t\t<gen:value>2016-12-22T00:00:00</gen:value>\n",
    "\t\t\t</gen:stringAttribute>\n",
    "\t\t\t<gen:stringAttribute name=\"zrodloDach\">\n",
    "\t\t\t\t<gen:value>ALS_I</gen:value>\n",
    "\t\t\t</gen:stringAttribute>\n",
    "\t\t\t<gen:stringAttribute name=\"aktZrodla\">\n",
    "\t\t\t\t<gen:value>2011</gen:value>\n",
    "\t\t\t</gen:stringAttribute>\n",
    "\t\t\t<bldg:roofType>1000</bldg:roofType>\n",
    "\t\t\t<bldg:lod1Solid>\n",
    "\t\t\t\t<gml:Solid srsName=\"urn:ogc:def:crs,crs:EPSG::2180,crs:PL-KRON86-NH\" srsDimension=\"3\">\n",
    "\t\t\t\t\t<gml:exterior>\n",
    "\t\t\t\t\t\t<gml:CompositeSurface>\n",
    "\t\t\t\t\t\t\t<gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t\t<gml:Polygon gml:id=\"ID-1432-38E036F4-D4A5-9AC7-E053-CC2BA8C036B5_PG_1\">\n",
    "\t\t\t\t\t\t\t\t\t<gml:exterior>\n",
    "\t\t\t\t\t\t\t\t\t\t<gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t\t\t<gml:posList>624092.85 499386.21 78.75 624088.71 499373.86 78.75 624079.71 499376.87 78.75 624083.85 499389.22 78.75 624092.85 499386.21 78.75</gml:posList>\n",
    "\t\t\t\t\t\t\t\t\t\t</gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t</gml:exterior>\n",
    "\t\t\t\t\t\t\t\t</gml:Polygon>\n",
    "\t\t\t\t\t\t\t</gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t<gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t\t<gml:Polygon gml:id=\"ID-1432-38E036F4-D4A5-9AC7-E053-CC2BA8C036B5_PG_2\">\n",
    "\t\t\t\t\t\t\t\t\t<gml:exterior>\n",
    "\t\t\t\t\t\t\t\t\t\t<gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t\t\t<gml:posList>624092.85 499386.21 78.75 624083.85 499389.22 78.75 624083.85 499389.22 84.85000000000001 624092.85 499386.21 84.85000000000001 624092.85 499386.21 78.75</gml:posList>\n",
    "\t\t\t\t\t\t\t\t\t\t</gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t</gml:exterior>\n",
    "\t\t\t\t\t\t\t\t</gml:Polygon>\n",
    "\t\t\t\t\t\t\t</gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t<gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t\t<gml:Polygon gml:id=\"ID-1432-38E036F4-D4A5-9AC7-E053-CC2BA8C036B5_PG_3\">\n",
    "\t\t\t\t\t\t\t\t\t<gml:exterior>\n",
    "\t\t\t\t\t\t\t\t\t\t<gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t\t\t<gml:posList>624083.85 499389.22 78.75 624079.71 499376.87 78.75 624079.71 499376.87 84.85000000000001 624083.85 499389.22 84.85000000000001 624083.85 499389.22 78.75</gml:posList>\n",
    "\t\t\t\t\t\t\t\t\t\t</gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t</gml:exterior>\n",
    "\t\t\t\t\t\t\t\t</gml:Polygon>\n",
    "\t\t\t\t\t\t\t</gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t<gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t\t<gml:Polygon gml:id=\"ID-1432-38E036F4-D4A5-9AC7-E053-CC2BA8C036B5_PG_4\">\n",
    "\t\t\t\t\t\t\t\t\t<gml:exterior>\n",
    "\t\t\t\t\t\t\t\t\t\t<gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t\t\t<gml:posList>624079.71 499376.87 78.75 624088.71 499373.86 78.75 624088.71 499373.86 84.85000000000001 624079.71 499376.87 84.85000000000001 624079.71 499376.87 78.75</gml:posList>\n",
    "\t\t\t\t\t\t\t\t\t\t</gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t</gml:exterior>\n",
    "\t\t\t\t\t\t\t\t</gml:Polygon>\n",
    "\t\t\t\t\t\t\t</gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t<gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t\t<gml:Polygon gml:id=\"ID-1432-38E036F4-D4A5-9AC7-E053-CC2BA8C036B5_PG_5\">\n",
    "\t\t\t\t\t\t\t\t\t<gml:exterior>\n",
    "\t\t\t\t\t\t\t\t\t\t<gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t\t\t<gml:posList>624088.71 499373.86 78.75 624092.85 499386.21 78.75 624092.85 499386.21 84.85000000000001 624088.71 499373.86 84.85000000000001 624088.71 499373.86 78.75</gml:posList>\n",
    "\t\t\t\t\t\t\t\t\t\t</gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t</gml:exterior>\n",
    "\t\t\t\t\t\t\t\t</gml:Polygon>\n",
    "\t\t\t\t\t\t\t</gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t<gml:surfaceMember>\n",
    "\t\t\t\t\t\t\t\t<gml:Polygon gml:id=\"ID-1432-38E036F4-D4A5-9AC7-E053-CC2BA8C036B5_PG_6\">\n",
    "\t\t\t\t\t\t\t\t\t<gml:exterior>\n",
    "\t\t\t\t\t\t\t\t\t\t<gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t\t\t<gml:posList>624092.85 499386.21 84.85000000000001 624083.85 499389.22 84.85000000000001 624079.71 499376.87 84.85000000000001 624088.71 499373.86 84.85000000000001 624092.85 499386.21 84.85000000000001</gml:posList>\n",
    "\t\t\t\t\t\t\t\t\t\t</gml:LinearRing>\n",
    "\t\t\t\t\t\t\t\t\t</gml:exterior>\n",
    "\t\t\t\t\t\t\t\t</gml:Polygon>\n",
    "\t\t\t\t\t\t\t</gml:surfaceMember>\n",
    "\t\t\t\t\t\t</gml:CompositeSurface>\n",
    "\t\t\t\t\t</gml:exterior>\n",
    "\t\t\t\t</gml:Solid>\n",
    "\t\t\t</bldg:lod1Solid>\n",
    "\t\t</bldg:Building>\n",
    "\t</core:cityObjectMember>\n",
    "```\n",
    "\n",
    "FME było w stanie odczytać zipowane pliki, dokonać odpowiednich transformacji tj. \"spłaszczyć\" geometrię do 2D i wybrać część poligonową, a następnie załadować te dane do bazy PostgreSQL z dodatkiem PostGIS. \n",
    "\n",
    "Dane są co prawda sprzed kilku lat, ale jest to najpełniejsze źródło jakie znalazłem. Innym dobrym źródłem dla obrysów budynków jest OpenStreetMap (OSM), jednak o ile aktuaność danych OSM jest bardzo dobra, bo są aktualizowane na bieżąco to nie ma tam danych dla całego krajego, a próba łączenia dwóch zbiorów tej wielkości wykraczała by poza zakres tej pracy.\n",
    "\n",
    "Pliki workspace FME są również dostępne w reporzytorium git.\n",
    "\n",
    "### Przygotowanie danych z Numerycznego Modelu Terenu\n",
    "\n",
    "Dane są w formacie tekstowym, który można określić jako CSV rozdzielany spacjami. Każdy rekord zawiera 3 kolumny: x, y i z, czyli wysokość. Wersja darmowa tego zbioru ma punkty pomiarowe co 100m. Wysokość jest dla samego terenu (bez pokrycia jak budynki, drzewa itd.). W sumie dla całej Polski dało to ok. 31,2 mln punktów.\n",
    "\n",
    "Pierwsze 5 linijek z pliku:\n",
    "```\n",
    "222500.00 396100.00 162.20\n",
    "220300.00 396200.00 146.41\n",
    "220400.00 396200.00 149.84\n",
    "220500.00 396200.00 152.26\n",
    "220600.00 396200.00 151.81\n",
    "```\n",
    "\n",
    "Kolumny kolejno zawierają współrzędne X, Y i Z (wysokość). Współrzędne X i Y są w [układzie współrzędnych o kodzie EPSG 2180](https://pl.wikipedia.org/wiki/Uk%C5%82ad_wsp%C3%B3%C5%82rz%C4%99dnych_1992) potocznie zwanym CS92 lub PUWG1992.\n",
    "\n",
    "Dla każdego adresu wyszukamy punkty w odległości do 10km, a następnie policzymy minimalną wysokość, maksymalną oraz medianę wysokości. To da nam podstawowe dane na temat terenu na jakim znajduje się dany punkt.\n",
    "\n",
    "NMT można przetworzyć np. na raster i mogło by to uprościć obliczenia, ale na potrzeby tej pracy uznałem, że ta prosta metoda wystarczy. Tutaj też użyto bazy danych PostgreSQL z dodatkiem PostGIS. Co prawda rozmiar danych wymusił przetwarzanie ich małymi partiami, do czego napisany został skrypt w Pythonie, który wysyłał do bazy 8 takich kawałków jednocześnie (8, bo tyle rdzeni miała baza danych) pracując na 8 procesach stworzonych z biblioteką _multiprocessing_. Skrypt jest załączony w repozytorium git.\n",
    "\n",
    "### Połączenie danych w spójny słownik adresów wraz z dodatkowymi informacjami\n",
    "\n",
    "Mając przygotowane zbiory PRG, budynków oraz NMT możemy je połączyć, żeby uzyskać adres wraz z informacjami o tym ile adresów z nim sąsiaduje, do jakiej wielkości budynku należy oraz na jakim terenie się znajduje.\n",
    "\n",
    "Utworzymy kilka tabel, które potem połączymy w całość:\n",
    "- tabela z liczbą sąsiadów dla adresu - łączymy tabelę z wyczyszczonymi i zdeduplikowanymi adresami z PRG samą ze sobą z jednym warunkiem: szukamy sąsiadów w maksymalnej odległości 500m:\n",
    "\n",
    "```sql\n",
    "create table neighbours as \n",
    "    SELECT \n",
    "      a.lokalnyid, \n",
    "      count(*)-1 no_of_neighbours\n",
    "      -- odejmujemy jeden ponieważ łączymy tabelę samą do siebie\n",
    "      -- więc każdy punkt będzie miał przynajmniej jednego sąsiada: siebie\n",
    "    FROM dedup a\n",
    "    JOIN dedup b ON ST_DWithin(a.gml, b.gml, 500)\n",
    "    GROUP BY a.lokalnyid\n",
    ";\n",
    "```\n",
    "\n",
    "- tabela z powierzchnią najbliższego budynku - łączymy tabelę adresów z tabelą obrysów budyków z warunkiem maksymalnej odległości 50m, następnie numerujemy znalezione budynki dla każdego adresu po odległości między nimi i wybieramy ten najbliższy:\n",
    "\n",
    "```sql\n",
    "create table areas as \n",
    "    select lokalnyid, st_area(geom) building_area\n",
    "    from (\n",
    "        -- znajdź dla każdego punktu adresowego budynki oddalone maks o 50m\n",
    "        -- następnie je ponumeruj od najbliższych do najdalszych\n",
    "        select\n",
    "            a.lokalnyid,\n",
    "            b.geom,\n",
    "            row_number() over(partition by a.lokalnyid order by ST_Distance(a.gml, b.geom)) rn\n",
    "        from dedup a\n",
    "        join lod1_buildings_all b on ST_DWithin(a.gml, b.geom, 50)\n",
    "    ) a\n",
    "    where rn = 1\n",
    ";\n",
    "```\n",
    "\n",
    "- tworzymy tabelę na informacje o ukształtowaniu terenu - wypełnienie z użyciem skryptu _terrain_prepare.py_ w folderze skrypty:\n",
    "\n",
    "```sql\n",
    "create table terrain (\n",
    "    lokalnyid uuid primary key,\n",
    "    min_z double precision,\n",
    "    max_z double precision,\n",
    "    median_z double precision\n",
    ");\n",
    "```\n",
    "\n",
    "- ładujemy do bazy danych wcześniej przygotowane pliki CSV z danymi o zasięgach i węzłach, następnie uzupełniamy informacje o lokalizacji dla węzłów, grupujemy je po geometrii (lokalizacji) i agregujemy informacje o technologiach:\n",
    "\n",
    "\n",
    "```sql\n",
    "create table geonodes as\n",
    "    with a as (\n",
    "        select \n",
    "            teryt_simc\n",
    "            , coalesce(teryt_ulic, '99999') teryt_ulic\n",
    "            ,nr\n",
    "            ,x ,y\n",
    "        from wynik\n",
    "    ),\n",
    "    b as (\n",
    "        select a.x, a.y, n.fiber, n.copper, n.xdsl, n.radio\n",
    "        from nodes n\n",
    "        join a \n",
    "            on a.teryt_simc = n.city_code_value \n",
    "            and a.teryt_ulic = coalesce(n.street_code_value, '99999') \n",
    "            and a.nr = n.house_no\n",
    "        where latitude is null\n",
    "    ),\n",
    "    c as (\n",
    "        select n.longitude x, n.latitude y, n.fiber, n.copper, n.xdsl, n.radio\n",
    "        from nodes n\n",
    "        where latitude is not null\n",
    "    ),\n",
    "    z as (\n",
    "        select * from b\n",
    "        union all\n",
    "        select * from c\n",
    "    )\n",
    "    select \n",
    "        ST_SetSRID(ST_MakePoint(x, y),4326)::geography as geom,\n",
    "        max(fiber) fiber, \n",
    "        max(copper) copper, \n",
    "        max(xdsl) xdsl, \n",
    "        max(radio) radio\n",
    "    from z\n",
    "    group by ST_SetSRID(ST_MakePoint(x, y),4326)::geography\n",
    ";\n",
    "-- utworznie indeksu przestrzennego\n",
    "create index geonodes_geom_idx on geonodes using gist (geom);\n",
    "```\n",
    "\n",
    "- dla łatwiejszego liczenia tworzymy tabele tymczasowe z węzłami konkretnej technologii:\n",
    "\n",
    "```sql\n",
    "create unlogged table geonodes_fiber as \n",
    "    select geom::geography from geonodes where fiber = 1;\n",
    "create index on geonodes_fiber using gist (geom);\n",
    "\n",
    "create unlogged table geonodes_copper_xdsl as \n",
    "    select geom::geography from geonodes where copper = 1 and xdsl = 1;\n",
    "create index on geonodes_copper_xdsl using gist (geom);\n",
    "\n",
    "create unlogged table geonodes_copper_nonxdsl as \n",
    "    select geom::geography from geonodes where copper = 1 and xdsl = 0;\n",
    "create index on geonodes_copper_nonxdsl using gist (geom);\n",
    "\n",
    "create unlogged table geonodes_radio as \n",
    "    select geom::geography from geonodes where radio = 1;\n",
    "create index on geonodes_radio using gist (geom);\n",
    "```\n",
    "\n",
    "- w FME liczymy odległości od węzłów oraz ile nich sąsiaduje z adresami.\n",
    "\n",
    "- dalej łączymy to wszystko w jedną tabelę:\n",
    "\n",
    "```sql\n",
    "create table lookup_addr as\n",
    "    SELECT *\n",
    "    FROM geowynik w\n",
    "    LEFT JOIN temp_fiber_count using(lokalnyid)\n",
    "    LEFT JOIN temp_nonxdsl_count using(lokalnyid)\n",
    "    LEFT JOIN temp_radio_count using(lokalnyid)\n",
    "    LEFT JOIN temp_xdsl_count using(lokalnyid)\n",
    "    LEFT JOIN fiber using(lokalnyid)\n",
    "    LEFT JOIN radio using(lokalnyid)\n",
    "    LEFT JOIN copper_xdsl using(lokalnyid)\n",
    "    LEFT JOIN copper_nonxdsl using(lokalnyid)\n",
    ";\n",
    "```\n",
    "\n",
    "- na koniec tworzymy tabelę ze zbiorem po połączeniu danych o zasięgach i wcześniej przygotowanych danych dla punktów adresowych z PRG:\n",
    "\n",
    "```sql\n",
    "create table dataset as \n",
    "    with \n",
    "    -- punkty adresowe z zasięgów dla któych mamy dane adresowe\n",
    "    -- zostawiamy od razu tylko największą dostępną przepustowość\n",
    "    ep_adr as (\n",
    "        select \n",
    "            city_code_value teryt_simc, \n",
    "            coalesce(street_code_value) teryt_ulic, \n",
    "            house_no nr, \n",
    "            max(stationary_bandwidth_name) przepustowosc\n",
    "        from endpoints e\n",
    "        where house_no is not null\n",
    "        group by 1,2,3\n",
    "    ),\n",
    "    -- punkty adresowe z zasięgów dla których nie mamy adresu, ale mamy współrzędne\n",
    "    -- j.w.\n",
    "    ep_latlon as (\n",
    "        select \n",
    "            st_setsrid(st_makepoint(longitude, latitude), 4326) geom,\n",
    "            max(stationary_bandwidth_name) przepustowosc\n",
    "        from endpoints e\n",
    "        where latitude is not null and house_no is null\n",
    "        group by 1\n",
    "    )\n",
    "\n",
    "    select \n",
    "        ea.przepustowosc,\n",
    "        no_of_neighbours,\n",
    "        building_area,\n",
    "        delta_z,\n",
    "        median_z,\n",
    "        fiber_count,\n",
    "        copper_nonxdsl_count,\n",
    "        radio_count,\n",
    "        copper_xdsl_count,\n",
    "        distance_fiber,\n",
    "        distance_radio,\n",
    "        distance_copper_xdsl,\n",
    "        distance_copper_nonxdsl\n",
    "    from ep_adr ea\n",
    "    join lookup_addr la using(teryt_simc, teryt_ulic, nr)\n",
    "    UNION ALL\n",
    "    select \n",
    "        ell.przepustowosc,\n",
    "        no_of_neighbours,\n",
    "        building_area,\n",
    "        delta_z,\n",
    "        median_z,\n",
    "        fiber_count,\n",
    "        copper_nonxdsl_count,\n",
    "        radio_count,\n",
    "        copper_xdsl_count,\n",
    "        distance_fiber,\n",
    "        distance_radio,\n",
    "        distance_copper_xdsl,\n",
    "        distance_copper_nonxdsl\n",
    "    from ep_latlon ell\n",
    "    cross join lateral (\n",
    "        select *\n",
    "        from lookup_addr la\n",
    "        order by st_transform(la.geom, 2180) <-> st_transform(ell.geom, 2180)\n",
    "        limit 1\n",
    "    ) lgeo\n",
    ";\n",
    "```\n",
    "\n",
    "- tabelę eksportujemy z bazy do pliku CSV, który zaraz wczytamy do jupyter notebooka\n",
    "\n",
    "## Przygotowanie kompletnego zbioru do uczenia modelu\n",
    "\n",
    "Po operacjach z wczesniejszych działów otrzymujemy zbiór jak poniżej:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>przepustowosc</th>\n",
       "      <th>no_of_neighbours</th>\n",
       "      <th>building_area</th>\n",
       "      <th>delta_z</th>\n",
       "      <th>median_z</th>\n",
       "      <th>fiber_count</th>\n",
       "      <th>copper_nonxdsl_count</th>\n",
       "      <th>radio_count</th>\n",
       "      <th>copper_xdsl_count</th>\n",
       "      <th>distance_fiber</th>\n",
       "      <th>distance_radio</th>\n",
       "      <th>distance_copper_xdsl</th>\n",
       "      <th>distance_copper_nonxdsl</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>40.0</td>\n",
       "      <td>127</td>\n",
       "      <td>132.74060</td>\n",
       "      <td>135.619995</td>\n",
       "      <td>205.910004</td>\n",
       "      <td>131</td>\n",
       "      <td>459</td>\n",
       "      <td>37</td>\n",
       "      <td>12</td>\n",
       "      <td>2139.31</td>\n",
       "      <td>90.62</td>\n",
       "      <td>2139.31</td>\n",
       "      <td>2487.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8.0</td>\n",
       "      <td>293</td>\n",
       "      <td>144.79525</td>\n",
       "      <td>33.070000</td>\n",
       "      <td>100.870003</td>\n",
       "      <td>67</td>\n",
       "      <td>151</td>\n",
       "      <td>58</td>\n",
       "      <td>33</td>\n",
       "      <td>759.25</td>\n",
       "      <td>549.57</td>\n",
       "      <td>985.47</td>\n",
       "      <td>616.53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>60.0</td>\n",
       "      <td>166</td>\n",
       "      <td>87.74550</td>\n",
       "      <td>144.829987</td>\n",
       "      <td>237.369995</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>11</td>\n",
       "      <td>2</td>\n",
       "      <td>389.73</td>\n",
       "      <td>346.92</td>\n",
       "      <td>389.73</td>\n",
       "      <td>3072.64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>60.0</td>\n",
       "      <td>231</td>\n",
       "      <td>114.72625</td>\n",
       "      <td>239.630005</td>\n",
       "      <td>210.169998</td>\n",
       "      <td>9</td>\n",
       "      <td>50</td>\n",
       "      <td>78</td>\n",
       "      <td>25</td>\n",
       "      <td>532.56</td>\n",
       "      <td>486.90</td>\n",
       "      <td>522.59</td>\n",
       "      <td>511.16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>30.0</td>\n",
       "      <td>384</td>\n",
       "      <td>661.19260</td>\n",
       "      <td>466.490021</td>\n",
       "      <td>380.785004</td>\n",
       "      <td>293</td>\n",
       "      <td>211</td>\n",
       "      <td>79</td>\n",
       "      <td>46</td>\n",
       "      <td>101.62</td>\n",
       "      <td>132.88</td>\n",
       "      <td>170.08</td>\n",
       "      <td>125.94</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   przepustowosc  no_of_neighbours  building_area     delta_z    median_z  \\\n",
       "0           40.0               127      132.74060  135.619995  205.910004   \n",
       "1            8.0               293      144.79525   33.070000  100.870003   \n",
       "2           60.0               166       87.74550  144.829987  237.369995   \n",
       "3           60.0               231      114.72625  239.630005  210.169998   \n",
       "4           30.0               384      661.19260  466.490021  380.785004   \n",
       "\n",
       "   fiber_count  copper_nonxdsl_count  radio_count  copper_xdsl_count  \\\n",
       "0          131                   459           37                 12   \n",
       "1           67                   151           58                 33   \n",
       "2            2                     2           11                  2   \n",
       "3            9                    50           78                 25   \n",
       "4          293                   211           79                 46   \n",
       "\n",
       "   distance_fiber  distance_radio  distance_copper_xdsl  \\\n",
       "0         2139.31           90.62               2139.31   \n",
       "1          759.25          549.57                985.47   \n",
       "2          389.73          346.92                389.73   \n",
       "3          532.56          486.90                522.59   \n",
       "4          101.62          132.88                170.08   \n",
       "\n",
       "   distance_copper_nonxdsl  \n",
       "0                  2487.99  \n",
       "1                   616.53  \n",
       "2                  3072.64  \n",
       "3                   511.16  \n",
       "4                   125.94  "
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "df = pd.read_csv(r'dataset.zip', sep=';', compression='zip')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Kolumny oznaczają kolejno:\n",
    "* przepustowosc - maksymalna dostępna przepustowość internetu dla tego miejsca (mbit/s)\n",
    "* no_of_neighbours - liczba sąsiadów adresu w odległości maksymalnie 500 m\n",
    "* building_area - powierzchnia najbliższego budynku w m2\n",
    "* delta_z - zmiana wysokości terenu (dla punktów pomiarowych w odległości maksymalnie 10 km)\n",
    "* median_z - mediana wysokości terenu (z punktów pomiarowych w odległości maksymalnie 10 km)\n",
    "* fiber_count - liczba węzłów światłowodowych w odległości maksymalnie 5 km\n",
    "* copper_nonxdsl_count - liczba węzłów w technologii miedzianej nie bazowanej na xDSL w odległości maksymalnie 5 km\n",
    "* radio_count - liczba węzłów radiowych w odległości maksymalnie 5 km\n",
    "* copper_xdsl_count - liczba węzłów w technologii xDSL w odległości maksymalnie 5 km\n",
    "* distance_fiber - odległość od najbliższego węzła światłowodowego (maksymalnie 20km lub oznaczony jako brak danych)\n",
    "* distance_radio - odległość od najbliższego węzła radiowego (j.w.)\n",
    "* distance_copper_xdsl - odległość od najbliższego węzła w technologii xDSL (j.w.)\n",
    "* distance_copper_nonxdsl - odległość od nabliższego węzła w technologii miedzianej nie bazowanej na xDSL (j.w.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Zbiór posiada: 5415128 rekordów.\n"
     ]
    }
   ],
   "source": [
    "print('Zbiór posiada:', len(df.index), 'rekordów.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\"Kształt\" naszego DataFrame:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5415128, 13)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Typy kolumn to:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "przepustowosc              float64\n",
       "no_of_neighbours             int64\n",
       "building_area              float64\n",
       "delta_z                    float64\n",
       "median_z                   float64\n",
       "fiber_count                  int64\n",
       "copper_nonxdsl_count         int64\n",
       "radio_count                  int64\n",
       "copper_xdsl_count            int64\n",
       "distance_fiber             float64\n",
       "distance_radio             float64\n",
       "distance_copper_xdsl       float64\n",
       "distance_copper_nonxdsl    float64\n",
       "dtype: object"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.dtypes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Podstawowe informacje na temat zawartości:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>przepustowosc</th>\n",
       "      <th>no_of_neighbours</th>\n",
       "      <th>building_area</th>\n",
       "      <th>delta_z</th>\n",
       "      <th>median_z</th>\n",
       "      <th>fiber_count</th>\n",
       "      <th>copper_nonxdsl_count</th>\n",
       "      <th>radio_count</th>\n",
       "      <th>copper_xdsl_count</th>\n",
       "      <th>distance_fiber</th>\n",
       "      <th>distance_radio</th>\n",
       "      <th>distance_copper_xdsl</th>\n",
       "      <th>distance_copper_nonxdsl</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5371957.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5414891.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5415128.00</td>\n",
       "      <td>5413571.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>418.90</td>\n",
       "      <td>213.61</td>\n",
       "      <td>167.14</td>\n",
       "      <td>154.77</td>\n",
       "      <td>186.51</td>\n",
       "      <td>137.15</td>\n",
       "      <td>262.36</td>\n",
       "      <td>75.56</td>\n",
       "      <td>21.50</td>\n",
       "      <td>1667.60</td>\n",
       "      <td>706.44</td>\n",
       "      <td>1076.33</td>\n",
       "      <td>2059.85</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>3672.18</td>\n",
       "      <td>188.35</td>\n",
       "      <td>434.18</td>\n",
       "      <td>179.43</td>\n",
       "      <td>118.55</td>\n",
       "      <td>297.38</td>\n",
       "      <td>662.75</td>\n",
       "      <td>118.04</td>\n",
       "      <td>33.96</td>\n",
       "      <td>2173.94</td>\n",
       "      <td>843.25</td>\n",
       "      <td>1077.62</td>\n",
       "      <td>2766.39</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>5.35</td>\n",
       "      <td>6.04</td>\n",
       "      <td>-0.44</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>10.00</td>\n",
       "      <td>61.00</td>\n",
       "      <td>88.25</td>\n",
       "      <td>66.55</td>\n",
       "      <td>103.87</td>\n",
       "      <td>3.00</td>\n",
       "      <td>2.00</td>\n",
       "      <td>15.00</td>\n",
       "      <td>4.00</td>\n",
       "      <td>254.48</td>\n",
       "      <td>204.96</td>\n",
       "      <td>332.61</td>\n",
       "      <td>240.30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>40.00</td>\n",
       "      <td>154.00</td>\n",
       "      <td>111.16</td>\n",
       "      <td>93.98</td>\n",
       "      <td>163.75</td>\n",
       "      <td>18.00</td>\n",
       "      <td>23.00</td>\n",
       "      <td>35.00</td>\n",
       "      <td>9.00</td>\n",
       "      <td>718.14</td>\n",
       "      <td>398.27</td>\n",
       "      <td>670.99</td>\n",
       "      <td>778.56</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>300.00</td>\n",
       "      <td>325.00</td>\n",
       "      <td>145.87</td>\n",
       "      <td>157.41</td>\n",
       "      <td>243.61</td>\n",
       "      <td>116.00</td>\n",
       "      <td>174.00</td>\n",
       "      <td>81.00</td>\n",
       "      <td>24.00</td>\n",
       "      <td>2315.39</td>\n",
       "      <td>828.31</td>\n",
       "      <td>1451.43</td>\n",
       "      <td>2900.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>100000.00</td>\n",
       "      <td>1286.00</td>\n",
       "      <td>151230.75</td>\n",
       "      <td>1596.34</td>\n",
       "      <td>1209.33</td>\n",
       "      <td>3118.00</td>\n",
       "      <td>6340.00</td>\n",
       "      <td>1445.00</td>\n",
       "      <td>360.00</td>\n",
       "      <td>19986.98</td>\n",
       "      <td>10390.25</td>\n",
       "      <td>12226.62</td>\n",
       "      <td>19999.88</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      przepustowosc no_of_neighbours building_area     delta_z    median_z  \\\n",
       "count    5415128.00       5415128.00    5371957.00  5415128.00  5415128.00   \n",
       "mean         418.90           213.61        167.14      154.77      186.51   \n",
       "std         3672.18           188.35        434.18      179.43      118.55   \n",
       "min            1.00             0.00          5.35        6.04       -0.44   \n",
       "25%           10.00            61.00         88.25       66.55      103.87   \n",
       "50%           40.00           154.00        111.16       93.98      163.75   \n",
       "75%          300.00           325.00        145.87      157.41      243.61   \n",
       "max       100000.00          1286.00     151230.75     1596.34     1209.33   \n",
       "\n",
       "      fiber_count copper_nonxdsl_count radio_count copper_xdsl_count  \\\n",
       "count  5415128.00           5415128.00  5415128.00        5415128.00   \n",
       "mean       137.15               262.36       75.56             21.50   \n",
       "std        297.38               662.75      118.04             33.96   \n",
       "min          0.00                 0.00        0.00              0.00   \n",
       "25%          3.00                 2.00       15.00              4.00   \n",
       "50%         18.00                23.00       35.00              9.00   \n",
       "75%        116.00               174.00       81.00             24.00   \n",
       "max       3118.00              6340.00     1445.00            360.00   \n",
       "\n",
       "      distance_fiber distance_radio distance_copper_xdsl  \\\n",
       "count     5414891.00     5415128.00           5415128.00   \n",
       "mean         1667.60         706.44              1076.33   \n",
       "std          2173.94         843.25              1077.62   \n",
       "min             0.00           0.00                 0.00   \n",
       "25%           254.48         204.96               332.61   \n",
       "50%           718.14         398.27               670.99   \n",
       "75%          2315.39         828.31              1451.43   \n",
       "max         19986.98       10390.25             12226.62   \n",
       "\n",
       "      distance_copper_nonxdsl  \n",
       "count              5413571.00  \n",
       "mean                  2059.85  \n",
       "std                   2766.39  \n",
       "min                      0.00  \n",
       "25%                    240.30  \n",
       "50%                    778.56  \n",
       "75%                   2900.20  \n",
       "max                  19999.88  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe().apply(lambda s: s.apply(lambda x: format(x, '.2f')))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sprawdźmy jeszcze ile rekordów nie ma danych:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "przepustowosc                  0\n",
       "no_of_neighbours               0\n",
       "building_area              43171\n",
       "delta_z                        0\n",
       "median_z                       0\n",
       "fiber_count                    0\n",
       "copper_nonxdsl_count           0\n",
       "radio_count                    0\n",
       "copper_xdsl_count              0\n",
       "distance_fiber               237\n",
       "distance_radio                 0\n",
       "distance_copper_xdsl           0\n",
       "distance_copper_nonxdsl     1557\n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.isnull().sum(axis = 0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Braków danych w tym zbiorze jest na tyle mało, że można je po prostu usunąć:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.dropna()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Dalej zajmiemy się wartościami krańcowymi, które mogły by zaburzyć działanie modelu.\n",
    "\n",
    "Zwizualizujemy zmienne przepustowość i liczba węzłów nie-xDSL na histogramie. Dodatkowo zwizualizujemy wartości powierzchni budynków.\n",
    "\n",
    "Dodatkowo kolumny z dystansem w metrach poddamy działaniu funkcji log(n+1). Logarytm \"spłaszczy\" nam różnice w odległości i sprawi, że nie będziemy już operować na olbrzymich wartościach liczbowych. Do odległości dodajemy 1 żeby nie było błędu dla logarytmu odległości równych zero."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1f4823a01f0>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df['przepustowosc'].groupby(df['przepustowosc']).count().plot(kind='barh')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6309"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp = df['copper_nonxdsl_count'].groupby(df['copper_nonxdsl_count']).count()\n",
    "len(temp.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "398080"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp[temp.index >= 1000].sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "copper_nonxdsl_count\n",
       "1000    397\n",
       "1001    365\n",
       "1002    383\n",
       "1003    360\n",
       "1004    369\n",
       "       ... \n",
       "6332      1\n",
       "6333      2\n",
       "6335      3\n",
       "6336      2\n",
       "6340      1\n",
       "Name: copper_nonxdsl_count, Length: 5309, dtype: int64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp[temp.index >= 1000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1f4824655b0>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "temp.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1f482530f10>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "temp[temp.index <= 100].plot()  # wykres dla liczby rekordów do 100 węzłów"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4789246    151230.746050\n",
       "3278193     97594.134850\n",
       "3518577     95780.454751\n",
       "1217863     87784.993750\n",
       "4634798     81330.268800\n",
       "               ...      \n",
       "3000080     11100.606950\n",
       "1733732     11099.611295\n",
       "1079710     11099.611295\n",
       "3826826     11094.068000\n",
       "4307215     11079.698850\n",
       "Name: building_area, Length: 1000, dtype: float64"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['building_area'].sort_values(ascending=False)[:1000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1f4825871f0>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df['building_area'].plot(kind='hist', bins=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1f4825b9280>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df[df['building_area'] < 600]['building_area'].hist(bins=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1175"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['building_area'] >= 10000]['building_area'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "17294"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[df['przepustowosc'] > 2500]['przepustowosc'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x864 with 16 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "_ = df.hist(figsize=(16,12), bins=20)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Przepustowości zostaną podzielone na klasy, które będą przewidywane przez model. Wyróżnimy 5 klas:\n",
    "\n",
    "1 - 1-8 mbit/s  \n",
    "2 - 10-20 mbit/s  \n",
    "3 - 30-80 mbit/s  \n",
    "4 - 100-600 mbit/s  \n",
    "5 - >=1000 mbit/s\n",
    "\n",
    "(klasy nie zazęiają się bo dane sprawozdawane do urzędu mają z góry określone wartości, tak więc nie ma możliwości przekazania przepustowości jak 90 mbit/s).\n",
    "\n",
    "Usuniemy rekordy dla powierzchni budynków powyżej 10000.\n",
    "\n",
    "Z liczbą węzłów ciężej ocenić, czy część z nich byłaby do usunięcia jako tzw. _outliery_, ponieważ jest bardo duży przedział wartości zawierających po kilka kilkadziesiąt rekordów. Pozostanie skalowanie dla zbioru."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df[df['building_area'] <= 10000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "classes = {\n",
    "    1: {'from': 0, 'to': 9, 'desc': '1-8 mbit/s'},\n",
    "    2: {'from': 10, 'to': 20, 'desc': '10-20 mbit/s'},\n",
    "    3: {'from': 30, 'to': 80, 'desc': '30-80 mbit/s'},\n",
    "    4: {'from': 100, 'to': 600, 'desc': '100-600 mbit/s'},\n",
    "    5: {'from': 601, 'to': 9999999, 'desc': '>=1000 mbit/s'},\n",
    "}\n",
    "labels = list(classes.keys())\n",
    "desc_labels = [x['desc'] for x in classes.values()]\n",
    "\n",
    "def classify(x):\n",
    "    for k, v in classes.items():\n",
    "        if v['from'] <= x <= v['to']:\n",
    "            return k\n",
    "\n",
    "df['class'] = df['przepustowosc'].apply(classify)\n",
    "df.drop(columns=['przepustowosc'], inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numpy import log1p\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "df['distance_copper_nonxdsl'] = df['distance_copper_nonxdsl'].apply(log1p)\n",
    "df['distance_copper_xdsl'] = df['distance_copper_xdsl'].apply(log1p)\n",
    "df['distance_fiber'] = df['distance_fiber'].apply(log1p)\n",
    "df['distance_radio'] = df['distance_radio'].apply(log1p)\n",
    "df['copper_nonxdsl_count'] = df['copper_nonxdsl_count'].apply(log1p)\n",
    "df['copper_xdsl_count'] = df['copper_xdsl_count'].apply(log1p)\n",
    "df['fiber_count'] = df['fiber_count'].apply(log1p)\n",
    "df['radio_count'] = df['radio_count'].apply(log1p)\n",
    "df['building_area'] = df['building_area'].apply(log1p)\n",
    "df['no_of_neighbours'] = df['no_of_neighbours'].apply(log1p)\n",
    "\n",
    "mms = MinMaxScaler()\n",
    "df[['delta_z','median_z']] = mms.fit_transform(df[['delta_z','median_z']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# change DataFrame column order\n",
    "cols = [\n",
    "    'class',\n",
    "    'building_area',\n",
    "    'no_of_neighbours',\n",
    "    'delta_z',\n",
    "    'median_z',\n",
    "    'distance_copper_nonxdsl',\n",
    "    'distance_copper_xdsl',\n",
    "    'distance_fiber',\n",
    "    'distance_radio',\n",
    "    'copper_nonxdsl_count',\n",
    "    'copper_xdsl_count',\n",
    "    'fiber_count',\n",
    "    'radio_count',\n",
    "]\n",
    "df = df[cols]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x864 with 16 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "_ = df.hist(figsize=(16,12), bins=20)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Spójrzemy teraz jak wyglądają nasze dane, które wykorzystamy do modelu."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>class</th>\n",
       "      <th>building_area</th>\n",
       "      <th>no_of_neighbours</th>\n",
       "      <th>delta_z</th>\n",
       "      <th>median_z</th>\n",
       "      <th>distance_copper_nonxdsl</th>\n",
       "      <th>distance_copper_xdsl</th>\n",
       "      <th>distance_fiber</th>\n",
       "      <th>distance_radio</th>\n",
       "      <th>copper_nonxdsl_count</th>\n",
       "      <th>copper_xdsl_count</th>\n",
       "      <th>fiber_count</th>\n",
       "      <th>radio_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>4.895902</td>\n",
       "      <td>4.852030</td>\n",
       "      <td>0.081481</td>\n",
       "      <td>0.170570</td>\n",
       "      <td>7.819632</td>\n",
       "      <td>7.668706</td>\n",
       "      <td>7.668706</td>\n",
       "      <td>4.517650</td>\n",
       "      <td>6.131226</td>\n",
       "      <td>2.564949</td>\n",
       "      <td>4.882802</td>\n",
       "      <td>3.637586</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>4.982203</td>\n",
       "      <td>5.683580</td>\n",
       "      <td>0.016997</td>\n",
       "      <td>0.083743</td>\n",
       "      <td>6.425728</td>\n",
       "      <td>6.894133</td>\n",
       "      <td>6.633647</td>\n",
       "      <td>6.310954</td>\n",
       "      <td>5.023881</td>\n",
       "      <td>3.526361</td>\n",
       "      <td>4.219508</td>\n",
       "      <td>4.077537</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>4.485773</td>\n",
       "      <td>5.117994</td>\n",
       "      <td>0.087273</td>\n",
       "      <td>0.196575</td>\n",
       "      <td>8.030618</td>\n",
       "      <td>5.968017</td>\n",
       "      <td>5.968017</td>\n",
       "      <td>5.851973</td>\n",
       "      <td>1.098612</td>\n",
       "      <td>1.098612</td>\n",
       "      <td>1.098612</td>\n",
       "      <td>2.484907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>4.751227</td>\n",
       "      <td>5.446737</td>\n",
       "      <td>0.146884</td>\n",
       "      <td>0.174091</td>\n",
       "      <td>6.238637</td>\n",
       "      <td>6.260709</td>\n",
       "      <td>6.279572</td>\n",
       "      <td>6.190110</td>\n",
       "      <td>3.931826</td>\n",
       "      <td>3.258097</td>\n",
       "      <td>2.302585</td>\n",
       "      <td>4.369448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>6.495556</td>\n",
       "      <td>5.953243</td>\n",
       "      <td>0.289537</td>\n",
       "      <td>0.315122</td>\n",
       "      <td>4.843715</td>\n",
       "      <td>5.142131</td>\n",
       "      <td>4.631033</td>\n",
       "      <td>4.896944</td>\n",
       "      <td>5.356586</td>\n",
       "      <td>3.850148</td>\n",
       "      <td>5.683580</td>\n",
       "      <td>4.382027</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   class  building_area  no_of_neighbours   delta_z  median_z  \\\n",
       "0      3       4.895902          4.852030  0.081481  0.170570   \n",
       "1      1       4.982203          5.683580  0.016997  0.083743   \n",
       "2      3       4.485773          5.117994  0.087273  0.196575   \n",
       "3      3       4.751227          5.446737  0.146884  0.174091   \n",
       "4      3       6.495556          5.953243  0.289537  0.315122   \n",
       "\n",
       "   distance_copper_nonxdsl  distance_copper_xdsl  distance_fiber  \\\n",
       "0                 7.819632              7.668706        7.668706   \n",
       "1                 6.425728              6.894133        6.633647   \n",
       "2                 8.030618              5.968017        5.968017   \n",
       "3                 6.238637              6.260709        6.279572   \n",
       "4                 4.843715              5.142131        4.631033   \n",
       "\n",
       "   distance_radio  copper_nonxdsl_count  copper_xdsl_count  fiber_count  \\\n",
       "0        4.517650              6.131226           2.564949     4.882802   \n",
       "1        6.310954              5.023881           3.526361     4.219508   \n",
       "2        5.851973              1.098612           1.098612     1.098612   \n",
       "3        6.190110              3.931826           3.258097     2.302585   \n",
       "4        4.896944              5.356586           3.850148     5.683580   \n",
       "\n",
       "   radio_count  \n",
       "0     3.637586  \n",
       "1     4.077537  \n",
       "2     2.484907  \n",
       "3     4.369448  \n",
       "4     4.382027  "
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5369004, 13)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>class</th>\n",
       "      <th>building_area</th>\n",
       "      <th>no_of_neighbours</th>\n",
       "      <th>delta_z</th>\n",
       "      <th>median_z</th>\n",
       "      <th>distance_copper_nonxdsl</th>\n",
       "      <th>distance_copper_xdsl</th>\n",
       "      <th>distance_fiber</th>\n",
       "      <th>distance_radio</th>\n",
       "      <th>copper_nonxdsl_count</th>\n",
       "      <th>copper_xdsl_count</th>\n",
       "      <th>fiber_count</th>\n",
       "      <th>radio_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "      <td>5369004.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.90</td>\n",
       "      <td>4.82</td>\n",
       "      <td>4.89</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.15</td>\n",
       "      <td>6.61</td>\n",
       "      <td>6.50</td>\n",
       "      <td>6.53</td>\n",
       "      <td>5.99</td>\n",
       "      <td>3.29</td>\n",
       "      <td>2.44</td>\n",
       "      <td>3.10</td>\n",
       "      <td>3.61</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.33</td>\n",
       "      <td>0.59</td>\n",
       "      <td>1.11</td>\n",
       "      <td>0.11</td>\n",
       "      <td>0.10</td>\n",
       "      <td>1.67</td>\n",
       "      <td>1.05</td>\n",
       "      <td>1.53</td>\n",
       "      <td>1.14</td>\n",
       "      <td>2.35</td>\n",
       "      <td>1.12</td>\n",
       "      <td>2.05</td>\n",
       "      <td>1.21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1.00</td>\n",
       "      <td>1.85</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>2.00</td>\n",
       "      <td>4.49</td>\n",
       "      <td>4.13</td>\n",
       "      <td>0.04</td>\n",
       "      <td>0.09</td>\n",
       "      <td>5.48</td>\n",
       "      <td>5.81</td>\n",
       "      <td>5.54</td>\n",
       "      <td>5.32</td>\n",
       "      <td>1.10</td>\n",
       "      <td>1.61</td>\n",
       "      <td>1.39</td>\n",
       "      <td>2.77</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>3.00</td>\n",
       "      <td>4.72</td>\n",
       "      <td>5.05</td>\n",
       "      <td>0.06</td>\n",
       "      <td>0.14</td>\n",
       "      <td>6.66</td>\n",
       "      <td>6.51</td>\n",
       "      <td>6.57</td>\n",
       "      <td>5.99</td>\n",
       "      <td>3.18</td>\n",
       "      <td>2.30</td>\n",
       "      <td>2.89</td>\n",
       "      <td>3.58</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>4.00</td>\n",
       "      <td>4.99</td>\n",
       "      <td>5.79</td>\n",
       "      <td>0.10</td>\n",
       "      <td>0.20</td>\n",
       "      <td>7.98</td>\n",
       "      <td>7.28</td>\n",
       "      <td>7.75</td>\n",
       "      <td>6.72</td>\n",
       "      <td>5.16</td>\n",
       "      <td>3.22</td>\n",
       "      <td>4.75</td>\n",
       "      <td>4.41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>5.00</td>\n",
       "      <td>9.21</td>\n",
       "      <td>7.16</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>9.90</td>\n",
       "      <td>9.41</td>\n",
       "      <td>9.90</td>\n",
       "      <td>9.25</td>\n",
       "      <td>8.75</td>\n",
       "      <td>5.89</td>\n",
       "      <td>8.05</td>\n",
       "      <td>7.28</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            class building_area no_of_neighbours     delta_z    median_z  \\\n",
       "count  5369004.00    5369004.00       5369004.00  5369004.00  5369004.00   \n",
       "mean         2.90          4.82             4.89        0.09        0.15   \n",
       "std          1.33          0.59             1.11        0.11        0.10   \n",
       "min          1.00          1.85             0.00        0.00        0.00   \n",
       "25%          2.00          4.49             4.13        0.04        0.09   \n",
       "50%          3.00          4.72             5.05        0.06        0.14   \n",
       "75%          4.00          4.99             5.79        0.10        0.20   \n",
       "max          5.00          9.21             7.16        1.00        1.00   \n",
       "\n",
       "      distance_copper_nonxdsl distance_copper_xdsl distance_fiber  \\\n",
       "count              5369004.00           5369004.00     5369004.00   \n",
       "mean                     6.61                 6.50           6.53   \n",
       "std                      1.67                 1.05           1.53   \n",
       "min                      0.00                 0.00           0.00   \n",
       "25%                      5.48                 5.81           5.54   \n",
       "50%                      6.66                 6.51           6.57   \n",
       "75%                      7.98                 7.28           7.75   \n",
       "max                      9.90                 9.41           9.90   \n",
       "\n",
       "      distance_radio copper_nonxdsl_count copper_xdsl_count fiber_count  \\\n",
       "count     5369004.00           5369004.00        5369004.00  5369004.00   \n",
       "mean            5.99                 3.29              2.44        3.10   \n",
       "std             1.14                 2.35              1.12        2.05   \n",
       "min             0.00                 0.00              0.00        0.00   \n",
       "25%             5.32                 1.10              1.61        1.39   \n",
       "50%             5.99                 3.18              2.30        2.89   \n",
       "75%             6.72                 5.16              3.22        4.75   \n",
       "max             9.25                 8.75              5.89        8.05   \n",
       "\n",
       "      radio_count  \n",
       "count  5369004.00  \n",
       "mean         3.61  \n",
       "std          1.21  \n",
       "min          0.00  \n",
       "25%          2.77  \n",
       "50%          3.58  \n",
       "75%          4.41  \n",
       "max          7.28  "
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe().apply(lambda s: s.apply(lambda x: format(x, '.2f')))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Liczebność klas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1f48495c7f0>"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df['class'].groupby(df['class']).count().plot(kind='bar')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Macierz korelacji."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style  type=\"text/css\" >\n",
       "    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col0 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col1 {\n",
       "            background-color:  #adc9fd;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col2 {\n",
       "            background-color:  #ead5c9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col3 {\n",
       "            background-color:  #445acc;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col4 {\n",
       "            background-color:  #6180e9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col5 {\n",
       "            background-color:  #98b9ff;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col6 {\n",
       "            background-color:  #6e90f2;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col7 {\n",
       "            background-color:  #7597f6;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col8 {\n",
       "            background-color:  #7396f5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col9 {\n",
       "            background-color:  #e7d7ce;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col10 {\n",
       "            background-color:  #dfdbd9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col11 {\n",
       "            background-color:  #e7d7ce;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col12 {\n",
       "            background-color:  #e2dad5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col0 {\n",
       "            background-color:  #b6cefa;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col1 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col2 {\n",
       "            background-color:  #c7d7f0;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col3 {\n",
       "            background-color:  #3f53c6;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col4 {\n",
       "            background-color:  #3f53c6;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col5 {\n",
       "            background-color:  #8fb1fe;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col6 {\n",
       "            background-color:  #89acfd;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col7 {\n",
       "            background-color:  #89acfd;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col8 {\n",
       "            background-color:  #7da0f9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col9 {\n",
       "            background-color:  #e5d8d1;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col10 {\n",
       "            background-color:  #dadce0;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col11 {\n",
       "            background-color:  #dadce0;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col12 {\n",
       "            background-color:  #dadce0;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col0 {\n",
       "            background-color:  #d4dbe6;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col1 {\n",
       "            background-color:  #98b9ff;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col2 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col3 {\n",
       "            background-color:  #3d50c3;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col4 {\n",
       "            background-color:  #3f53c6;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col5 {\n",
       "            background-color:  #5572df;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col6 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col7 {\n",
       "            background-color:  #4055c8;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col8 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col9 {\n",
       "            background-color:  #f39778;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col10 {\n",
       "            background-color:  #f6a385;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col11 {\n",
       "            background-color:  #f6a283;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col12 {\n",
       "            background-color:  #f6a385;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col0 {\n",
       "            background-color:  #90b2fe;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col1 {\n",
       "            background-color:  #80a3fa;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col2 {\n",
       "            background-color:  #b5cdfa;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col3 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col4 {\n",
       "            background-color:  #ec7f63;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col5 {\n",
       "            background-color:  #cad8ef;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col6 {\n",
       "            background-color:  #b2ccfb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col7 {\n",
       "            background-color:  #bfd3f6;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col8 {\n",
       "            background-color:  #9ebeff;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col9 {\n",
       "            background-color:  #c3d5f4;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col10 {\n",
       "            background-color:  #c0d4f5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col11 {\n",
       "            background-color:  #c1d4f4;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col12 {\n",
       "            background-color:  #ccd9ed;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col0 {\n",
       "            background-color:  #9dbdff;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col1 {\n",
       "            background-color:  #7295f4;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col2 {\n",
       "            background-color:  #abc8fd;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col3 {\n",
       "            background-color:  #ed8366;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col4 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col5 {\n",
       "            background-color:  #d6dce4;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col6 {\n",
       "            background-color:  #bbd1f8;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col7 {\n",
       "            background-color:  #c4d5f3;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col8 {\n",
       "            background-color:  #a3c2fe;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col9 {\n",
       "            background-color:  #b7cff9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col10 {\n",
       "            background-color:  #b7cff9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col11 {\n",
       "            background-color:  #bfd3f6;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col12 {\n",
       "            background-color:  #c5d6f2;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col0 {\n",
       "            background-color:  #536edd;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col1 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col2 {\n",
       "            background-color:  #3f53c6;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col3 {\n",
       "            background-color:  #485fd1;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col4 {\n",
       "            background-color:  #6e90f2;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col5 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col6 {\n",
       "            background-color:  #f5c1a9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col7 {\n",
       "            background-color:  #f39778;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col8 {\n",
       "            background-color:  #f7ba9f;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col9 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col10 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col11 {\n",
       "            background-color:  #4358cb;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col12 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col0 {\n",
       "            background-color:  #4257c9;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col1 {\n",
       "            background-color:  #516ddb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col2 {\n",
       "            background-color:  #4257c9;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col3 {\n",
       "            background-color:  #4257c9;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col4 {\n",
       "            background-color:  #5d7ce6;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col5 {\n",
       "            background-color:  #f7b89c;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col6 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col7 {\n",
       "            background-color:  #f7ad90;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col8 {\n",
       "            background-color:  #ebd3c6;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col9 {\n",
       "            background-color:  #84a7fc;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col10 {\n",
       "            background-color:  #5b7ae5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col11 {\n",
       "            background-color:  #7a9df8;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col12 {\n",
       "            background-color:  #799cf8;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col0 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col1 {\n",
       "            background-color:  #445acc;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col2 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col3 {\n",
       "            background-color:  #455cce;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col4 {\n",
       "            background-color:  #5b7ae5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col5 {\n",
       "            background-color:  #f29072;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col6 {\n",
       "            background-color:  #f7b194;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col7 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col8 {\n",
       "            background-color:  #f2cab5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col9 {\n",
       "            background-color:  #5e7de7;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col10 {\n",
       "            background-color:  #4a63d3;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col11 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col12 {\n",
       "            background-color:  #4e68d8;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col0 {\n",
       "            background-color:  #5977e3;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col1 {\n",
       "            background-color:  #5673e0;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col2 {\n",
       "            background-color:  #5572df;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col3 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col4 {\n",
       "            background-color:  #506bda;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col5 {\n",
       "            background-color:  #f7a98b;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col6 {\n",
       "            background-color:  #f0cdbb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col7 {\n",
       "            background-color:  #f6bea4;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col8 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col9 {\n",
       "            background-color:  #80a3fa;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col10 {\n",
       "            background-color:  #7396f5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col11 {\n",
       "            background-color:  #779af7;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col12 {\n",
       "            background-color:  #5a78e4;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col0 {\n",
       "            background-color:  #c3d5f4;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col1 {\n",
       "            background-color:  #b9d0f9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col2 {\n",
       "            background-color:  #f59f80;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col3 {\n",
       "            background-color:  #3c4ec2;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col4 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col5 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col6 {\n",
       "            background-color:  #6788ee;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col7 {\n",
       "            background-color:  #4f69d9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col8 {\n",
       "            background-color:  #516ddb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col9 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col10 {\n",
       "            background-color:  #d24b40;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col11 {\n",
       "            background-color:  #d55042;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col12 {\n",
       "            background-color:  #d1493f;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col0 {\n",
       "            background-color:  #c6d6f1;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col1 {\n",
       "            background-color:  #b7cff9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col2 {\n",
       "            background-color:  #f6a283;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col3 {\n",
       "            background-color:  #506bda;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col4 {\n",
       "            background-color:  #516ddb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col5 {\n",
       "            background-color:  #536edd;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col6 {\n",
       "            background-color:  #5875e1;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col7 {\n",
       "            background-color:  #5470de;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col8 {\n",
       "            background-color:  #5b7ae5;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col9 {\n",
       "            background-color:  #d0473d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col10 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col11 {\n",
       "            background-color:  #d95847;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col12 {\n",
       "            background-color:  #d44e41;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col0 {\n",
       "            background-color:  #ccd9ed;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col1 {\n",
       "            background-color:  #b1cbfc;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col2 {\n",
       "            background-color:  #f6a586;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col3 {\n",
       "            background-color:  #4961d2;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col4 {\n",
       "            background-color:  #5572df;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col5 {\n",
       "            background-color:  #516ddb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col6 {\n",
       "            background-color:  #6c8ff1;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col7 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col8 {\n",
       "            background-color:  #5673e0;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col9 {\n",
       "            background-color:  #d44e41;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col10 {\n",
       "            background-color:  #d95847;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col11 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col12 {\n",
       "            background-color:  #d75445;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col0 {\n",
       "            background-color:  #c6d6f1;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col1 {\n",
       "            background-color:  #b2ccfb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col2 {\n",
       "            background-color:  #f6a586;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col3 {\n",
       "            background-color:  #5e7de7;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col4 {\n",
       "            background-color:  #6180e9;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col5 {\n",
       "            background-color:  #4c66d6;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col6 {\n",
       "            background-color:  #6e90f2;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col7 {\n",
       "            background-color:  #516ddb;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col8 {\n",
       "            background-color:  #3c4ec2;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col9 {\n",
       "            background-color:  #d0473d;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col10 {\n",
       "            background-color:  #d55042;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col11 {\n",
       "            background-color:  #d75445;\n",
       "            color:  #000000;\n",
       "        }    #T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col12 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }</style><table id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86\" ><thead>    <tr>        <th class=\"blank level0\" ></th>        <th class=\"col_heading level0 col0\" >class</th>        <th class=\"col_heading level0 col1\" >building_area</th>        <th class=\"col_heading level0 col2\" >no_of_neighbours</th>        <th class=\"col_heading level0 col3\" >delta_z</th>        <th class=\"col_heading level0 col4\" >median_z</th>        <th class=\"col_heading level0 col5\" >distance_copper_nonxdsl</th>        <th class=\"col_heading level0 col6\" >distance_copper_xdsl</th>        <th class=\"col_heading level0 col7\" >distance_fiber</th>        <th class=\"col_heading level0 col8\" >distance_radio</th>        <th class=\"col_heading level0 col9\" >copper_nonxdsl_count</th>        <th class=\"col_heading level0 col10\" >copper_xdsl_count</th>        <th class=\"col_heading level0 col11\" >fiber_count</th>        <th class=\"col_heading level0 col12\" >radio_count</th>    </tr></thead><tbody>\n",
       "                <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row0\" class=\"row_heading level0 row0\" >class</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col0\" class=\"data row0 col0\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col1\" class=\"data row0 col1\" >0.15</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col2\" class=\"data row0 col2\" >0.28</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col3\" class=\"data row0 col3\" >0.0048</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col4\" class=\"data row0 col4\" >0.054</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col5\" class=\"data row0 col5\" >-0.24</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col6\" class=\"data row0 col6\" >-0.31</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col7\" class=\"data row0 col7\" >-0.34</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col8\" class=\"data row0 col8\" >-0.21</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col9\" class=\"data row0 col9\" >0.2</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col10\" class=\"data row0 col10\" >0.22</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col11\" class=\"data row0 col11\" >0.25</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row0_col12\" class=\"data row0 col12\" >0.22</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row1\" class=\"row_heading level0 row1\" >building_area</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col0\" class=\"data row1 col0\" >0.15</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col1\" class=\"data row1 col1\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col2\" class=\"data row1 col2\" >0.079</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col3\" class=\"data row1 col3\" >-0.0093</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col4\" class=\"data row1 col4\" >-0.061</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col5\" class=\"data row1 col5\" >-0.29</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col6\" class=\"data row1 col6\" >-0.19</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col7\" class=\"data row1 col7\" >-0.24</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col8\" class=\"data row1 col8\" >-0.17</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col9\" class=\"data row1 col9\" >0.2</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col10\" class=\"data row1 col10\" >0.19</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col11\" class=\"data row1 col11\" >0.17</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row1_col12\" class=\"data row1 col12\" >0.17</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row2\" class=\"row_heading level0 row2\" >no_of_neighbours</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col0\" class=\"data row2 col0\" >0.28</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col1\" class=\"data row2 col1\" >0.079</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col2\" class=\"data row2 col2\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col3\" class=\"data row2 col3\" >-0.017</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col4\" class=\"data row2 col4\" >-0.064</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col5\" class=\"data row2 col5\" >-0.58</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col6\" class=\"data row2 col6\" >-0.56</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col7\" class=\"data row2 col7\" >-0.6</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col8\" class=\"data row2 col8\" >-0.46</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col9\" class=\"data row2 col9\" >0.58</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col10\" class=\"data row2 col10\" >0.56</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col11\" class=\"data row2 col11\" >0.56</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row2_col12\" class=\"data row2 col12\" >0.55</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row3\" class=\"row_heading level0 row3\" >delta_z</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col0\" class=\"data row3 col0\" >0.0048</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col1\" class=\"data row3 col1\" >-0.0093</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col2\" class=\"data row3 col2\" >-0.017</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col3\" class=\"data row3 col3\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col4\" class=\"data row3 col4\" >0.8</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col5\" class=\"data row3 col5\" >0.02</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col6\" class=\"data row3 col6\" >-0.002</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col7\" class=\"data row3 col7\" >0.01</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col8\" class=\"data row3 col8\" >-0.028</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col9\" class=\"data row3 col9\" >-0.022</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col10\" class=\"data row3 col10\" >0.047</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col11\" class=\"data row3 col11\" >0.022</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row3_col12\" class=\"data row3 col12\" >0.093</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row4\" class=\"row_heading level0 row4\" >median_z</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col0\" class=\"data row4 col0\" >0.054</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col1\" class=\"data row4 col1\" >-0.061</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col2\" class=\"data row4 col2\" >-0.064</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col3\" class=\"data row4 col3\" >0.8</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col4\" class=\"data row4 col4\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col5\" class=\"data row4 col5\" >0.093</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col6\" class=\"data row4 col6\" >0.04</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col7\" class=\"data row4 col7\" >0.035</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col8\" class=\"data row4 col8\" >-0.0021</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col9\" class=\"data row4 col9\" >-0.081</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col10\" class=\"data row4 col10\" >0.0036</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col11\" class=\"data row4 col11\" >0.014</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row4_col12\" class=\"data row4 col12\" >0.054</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row5\" class=\"row_heading level0 row5\" >distance_copper_nonxdsl</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col0\" class=\"data row5 col0\" >-0.24</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col1\" class=\"data row5 col1\" >-0.29</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col2\" class=\"data row5 col2\" >-0.58</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col3\" class=\"data row5 col3\" >0.02</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col4\" class=\"data row5 col4\" >0.093</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col5\" class=\"data row5 col5\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col6\" class=\"data row5 col6\" >0.43</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col7\" class=\"data row5 col7\" >0.6</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col8\" class=\"data row5 col8\" >0.5</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col9\" class=\"data row5 col9\" >-0.73</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col10\" class=\"data row5 col10\" >-0.59</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col11\" class=\"data row5 col11\" >-0.59</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row5_col12\" class=\"data row5 col12\" >-0.62</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row6\" class=\"row_heading level0 row6\" >distance_copper_xdsl</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col0\" class=\"data row6 col0\" >-0.31</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col1\" class=\"data row6 col1\" >-0.19</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col2\" class=\"data row6 col2\" >-0.56</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col3\" class=\"data row6 col3\" >-0.002</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col4\" class=\"data row6 col4\" >0.04</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col5\" class=\"data row6 col5\" >0.43</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col6\" class=\"data row6 col6\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col7\" class=\"data row6 col7\" >0.51</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col8\" class=\"data row6 col8\" >0.36</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col9\" class=\"data row6 col9\" >-0.34</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col10\" class=\"data row6 col10\" >-0.42</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col11\" class=\"data row6 col11\" >-0.31</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row6_col12\" class=\"data row6 col12\" >-0.31</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row7\" class=\"row_heading level0 row7\" >distance_fiber</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col0\" class=\"data row7 col0\" >-0.34</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col1\" class=\"data row7 col1\" >-0.24</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col2\" class=\"data row7 col2\" >-0.6</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col3\" class=\"data row7 col3\" >0.01</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col4\" class=\"data row7 col4\" >0.035</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col5\" class=\"data row7 col5\" >0.6</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col6\" class=\"data row7 col6\" >0.51</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col7\" class=\"data row7 col7\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col8\" class=\"data row7 col8\" >0.42</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col9\" class=\"data row7 col9\" >-0.53</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col10\" class=\"data row7 col10\" >-0.5</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col11\" class=\"data row7 col11\" >-0.64</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row7_col12\" class=\"data row7 col12\" >-0.52</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row8\" class=\"row_heading level0 row8\" >distance_radio</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col0\" class=\"data row8 col0\" >-0.21</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col1\" class=\"data row8 col1\" >-0.17</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col2\" class=\"data row8 col2\" >-0.46</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col3\" class=\"data row8 col3\" >-0.028</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col4\" class=\"data row8 col4\" >-0.0021</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col5\" class=\"data row8 col5\" >0.5</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col6\" class=\"data row8 col6\" >0.36</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col7\" class=\"data row8 col7\" >0.42</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col8\" class=\"data row8 col8\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col9\" class=\"data row8 col9\" >-0.36</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col10\" class=\"data row8 col10\" >-0.31</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col11\" class=\"data row8 col11\" >-0.33</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row8_col12\" class=\"data row8 col12\" >-0.45</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row9\" class=\"row_heading level0 row9\" >copper_nonxdsl_count</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col0\" class=\"data row9 col0\" >0.2</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col1\" class=\"data row9 col1\" >0.2</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col2\" class=\"data row9 col2\" >0.58</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col3\" class=\"data row9 col3\" >-0.022</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col4\" class=\"data row9 col4\" >-0.081</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col5\" class=\"data row9 col5\" >-0.73</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col6\" class=\"data row9 col6\" >-0.34</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col7\" class=\"data row9 col7\" >-0.53</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col8\" class=\"data row9 col8\" >-0.36</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col9\" class=\"data row9 col9\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col10\" class=\"data row9 col10\" >0.86</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col11\" class=\"data row9 col11\" >0.84</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row9_col12\" class=\"data row9 col12\" >0.86</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row10\" class=\"row_heading level0 row10\" >copper_xdsl_count</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col0\" class=\"data row10 col0\" >0.22</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col1\" class=\"data row10 col1\" >0.19</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col2\" class=\"data row10 col2\" >0.56</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col3\" class=\"data row10 col3\" >0.047</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col4\" class=\"data row10 col4\" >0.0036</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col5\" class=\"data row10 col5\" >-0.59</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col6\" class=\"data row10 col6\" >-0.42</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col7\" class=\"data row10 col7\" >-0.5</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col8\" class=\"data row10 col8\" >-0.31</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col9\" class=\"data row10 col9\" >0.86</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col10\" class=\"data row10 col10\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col11\" class=\"data row10 col11\" >0.82</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row10_col12\" class=\"data row10 col12\" >0.85</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row11\" class=\"row_heading level0 row11\" >fiber_count</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col0\" class=\"data row11 col0\" >0.25</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col1\" class=\"data row11 col1\" >0.17</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col2\" class=\"data row11 col2\" >0.56</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col3\" class=\"data row11 col3\" >0.022</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col4\" class=\"data row11 col4\" >0.014</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col5\" class=\"data row11 col5\" >-0.59</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col6\" class=\"data row11 col6\" >-0.31</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col7\" class=\"data row11 col7\" >-0.64</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col8\" class=\"data row11 col8\" >-0.33</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col9\" class=\"data row11 col9\" >0.84</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col10\" class=\"data row11 col10\" >0.82</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col11\" class=\"data row11 col11\" >1</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row11_col12\" class=\"data row11 col12\" >0.83</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86level0_row12\" class=\"row_heading level0 row12\" >radio_count</th>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col0\" class=\"data row12 col0\" >0.22</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col1\" class=\"data row12 col1\" >0.17</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col2\" class=\"data row12 col2\" >0.55</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col3\" class=\"data row12 col3\" >0.093</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col4\" class=\"data row12 col4\" >0.054</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col5\" class=\"data row12 col5\" >-0.62</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col6\" class=\"data row12 col6\" >-0.31</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col7\" class=\"data row12 col7\" >-0.52</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col8\" class=\"data row12 col8\" >-0.45</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col9\" class=\"data row12 col9\" >0.86</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col10\" class=\"data row12 col10\" >0.85</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col11\" class=\"data row12 col11\" >0.83</td>\n",
       "                        <td id=\"T_66a42c17_3c94_11ea_8202_d8d09023ef86row12_col12\" class=\"data row12 col12\" >1</td>\n",
       "            </tr>\n",
       "    </tbody></table>"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x1f482c42250>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corr = df.corr()\n",
    "corr.style.background_gradient(cmap='coolwarm').set_precision(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Jak widać najbardziej skorelowanymi zmiennymi są odległości od węzła telekomunikacyjnego (odwrotna korelacja) w technologii światłowodowej i xDSL. Zmienne na temat ukształtowania terenu: zmiana wysokości i mediana wysokości nie wykazują prawie żadnej korelacji.\n",
    "\n",
    "Ostatnia rzecz to podzielenie danych na zbiór treningowy i testowy. Jako, że mamy dużo obserwacji zastosujemy podział 90% zbioru na trenowanie i 10% na testowanie."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "x = df[[x for x in df.columns if x != 'class']]\n",
    "y = df['class']\n",
    "\n",
    "train_x, test_x, train_y, test_y = train_test_split(x, y, train_size=0.9, test_size=0.1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model uczenia maszynowego\n",
    "\n",
    "Wykorzystamy algorytmy do klasyfikacji z biblioteki sklearn takie jak drzewo decyzyjne, K najbliższych sąsiadów, random forest oraz model gradientowy (HistGradientBoostingClassifier).\n",
    "\n",
    "Do oceny modeli użyjemy wskaźników accuracy score oraz balanced accuracy score, a także [confusion matrix](https://pl.wikipedia.org/wiki/Tablica_pomy%C5%82ek)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix, balanced_accuracy_score, plot_confusion_matrix, accuracy_score, classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.experimental import enable_hist_gradient_boosting\n",
    "from sklearn.ensemble import RandomForestClassifier, HistGradientBoostingClassifier\n",
    "\n",
    "models = {\n",
    "    'dt': DecisionTreeClassifier(),\n",
    "    'knn': KNeighborsClassifier(n_jobs=-1),\n",
    "    'rf': RandomForestClassifier(n_jobs=-1),\n",
    "    'xgb': HistGradientBoostingClassifier()  \n",
    "}\n",
    "predictions = {}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Trenujemy modele na naszych danych.\n",
    "\n",
    "komenda _%%time_ pokaże jak długo kod z danej komórki się wykonywał"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 5min 46s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "models['dt'].fit(train_x, train_y)\n",
    "predictions['dt'] = models['dt'].predict(test_x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 6min 39s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "models['knn'].fit(train_x, train_y)\n",
    "predictions['knn'] = models['knn'].predict(test_x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 3min 14s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "models['xgb'].fit(train_x, train_y)\n",
    "predictions['xgb'] = models['xgb'].predict(test_x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 12min\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "models['rf'].fit(train_x, train_y)\n",
    "predictions['rf'] = models['rf'].predict(test_x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Decision Tree\n",
      "balanced_accuracy_score: 0.7558776924836274\n",
      "accuracy_score: 0.7500395789912851\n",
      "classification_report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.74      0.74      0.74     93901\n",
      "           2       0.69      0.69      0.69    127691\n",
      "           3       0.75      0.75      0.75    142055\n",
      "           4       0.78      0.78      0.78     82134\n",
      "           5       0.83      0.82      0.82     91120\n",
      "\n",
      "    accuracy                           0.75    536901\n",
      "   macro avg       0.76      0.76      0.76    536901\n",
      "weighted avg       0.75      0.75      0.75    536901\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('model: Decision Tree')\n",
    "print('balanced_accuracy_score:', balanced_accuracy_score(test_y, predictions['dt']))\n",
    "print('accuracy_score:', accuracy_score(test_y, predictions['dt']))\n",
    "print('classification_report:')\n",
    "print(classification_report(test_y, predictions['dt'], labels=labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 69570,  12599,   6851,   2322,   2559],\n",
       "       [ 12807,  87642,  16714,   5914,   4614],\n",
       "       [  6583,  16623, 106424,   6858,   5567],\n",
       "       [  2335,   5774,   6850,  63994,   3181],\n",
       "       [  2561,   4621,   5660,   3211,  75067]], dtype=int64)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(test_y, predictions['dt'], labels=labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(10,10)\n",
    "_ = plot_confusion_matrix(models['dt'], test_x, test_y, labels, display_labels=desc_labels, normalize='pred', ax=ax)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: KNN\n",
      "balanced_accuracy_score: 0.7761879731390179\n",
      "accuracy_score: 0.7693187384638881\n",
      "classification_report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.74      0.81      0.77     93901\n",
      "           2       0.72      0.70      0.71    127691\n",
      "           3       0.78      0.76      0.77    142055\n",
      "           4       0.78      0.79      0.79     82134\n",
      "           5       0.84      0.82      0.83     91120\n",
      "\n",
      "    accuracy                           0.77    536901\n",
      "   macro avg       0.77      0.78      0.77    536901\n",
      "weighted avg       0.77      0.77      0.77    536901\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('model: KNN')\n",
    "print('balanced_accuracy_score:', balanced_accuracy_score(test_y, predictions['knn']))\n",
    "print('accuracy_score:', accuracy_score(test_y, predictions['knn']))\n",
    "print('classification_report:')\n",
    "print(classification_report(test_y, predictions['knn'], labels=labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 75649,   9550,   4198,   2234,   2270],\n",
       "       [ 13912,  89621,  14502,   5560,   4096],\n",
       "       [  7559,  15244, 107778,   6462,   5012],\n",
       "       [  2340,   5212,   6621,  65225,   2736],\n",
       "       [  2692,   4445,   5569,   3639,  74775]], dtype=int64)"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(test_y, predictions['knn'], labels=labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(10,10)\n",
    "_ = plot_confusion_matrix(models['knn'], test_x, test_y, labels, display_labels=desc_labels, normalize='pred', ax=ax)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: HistGradientBoosting\n",
      "balanced_accuracy_score: 0.5672331028087197\n",
      "accuracy_score: 0.5754040316557428\n",
      "classification_report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.58      0.66      0.62     93901\n",
      "           2       0.51      0.52      0.51    127691\n",
      "           3       0.58      0.66      0.62    142055\n",
      "           4       0.65      0.45      0.53     82134\n",
      "           5       0.62      0.54      0.58     91120\n",
      "\n",
      "    accuracy                           0.58    536901\n",
      "   macro avg       0.59      0.57      0.57    536901\n",
      "weighted avg       0.58      0.58      0.57    536901\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('model: HistGradientBoosting')\n",
    "print('balanced_accuracy_score:', balanced_accuracy_score(test_y, predictions['xgb']))\n",
    "print('accuracy_score:', accuracy_score(test_y, predictions['xgb']))\n",
    "print('classification_report:')\n",
    "print(classification_report(test_y, predictions['xgb'], labels=labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[62139, 18971,  6332,  1796,  4663],\n",
       "       [19209, 65912, 28280,  5934,  8356],\n",
       "       [10766, 20637, 94449,  6958,  9245],\n",
       "       [ 7049, 12338, 17404, 37288,  8055],\n",
       "       [ 8801, 12235, 15362,  5575, 49147]], dtype=int64)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(test_y, predictions['xgb'], labels=labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(10,10)\n",
    "_ = plot_confusion_matrix(models['xgb'], test_x, test_y, labels, display_labels=desc_labels, normalize='pred', ax=ax)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model: Random Forest\n",
      "balanced_accuracy_score: 0.82536873633716\n",
      "accuracy_score: 0.8183557117606411\n",
      "classification_report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.80      0.83      0.81     93901\n",
      "           2       0.78      0.75      0.77    127691\n",
      "           3       0.82      0.81      0.81    142055\n",
      "           4       0.83      0.85      0.84     82134\n",
      "           5       0.87      0.88      0.88     91120\n",
      "\n",
      "    accuracy                           0.82    536901\n",
      "   macro avg       0.82      0.83      0.82    536901\n",
      "weighted avg       0.82      0.82      0.82    536901\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('model: Random Forest')\n",
    "print('balanced_accuracy_score:', balanced_accuracy_score(test_y, predictions['rf']))\n",
    "print('accuracy_score:', accuracy_score(test_y, predictions['rf']))\n",
    "print('classification_report:')\n",
    "print(classification_report(test_y, predictions['rf'], labels=labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 77917,   8379,   3956,   1825,   1824],\n",
       "       [ 10841,  96230,  12357,   4792,   3471],\n",
       "       [  5957,  11740, 114753,   5400,   4205],\n",
       "       [  1556,   3605,   4743,  70158,   2072],\n",
       "       [  1713,   2981,   3814,   2294,  80318]], dtype=int64)"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(test_y, predictions['rf'], labels=labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(10,10)\n",
    "_ = plot_confusion_matrix(models['rf'], test_x, test_y, labels, display_labels=desc_labels, normalize='pred', ax=ax)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Jak widać najlepiej poradził sobie Random Forest, niedużo gorzej wypadły K najbliższych sąsiadów oraz pojedyncze drzewo decyzyjne. W zbiorze testowym były w stanie poprawnie przewidzieć ok. 70-80% przypadków to całkiem niezły wynik. MAcierz pokazuje nam także, że dla części klas modele radziły sobie lepiej dla części gorzej. Najłatwiej modelowi było zaklasyfikować rekordy dla najwyższych przepustowości, a najtrudniej dla najniższych. Dużo było przypadków gdzie model przewidział 1-8mbit/s, a w rzeczywistości było to 10-20mbit/s albo gdy model przewidział 30-80 mbit/s, a rzeczywistość to 10-20 mbit/s.\n",
    "\n",
    "## Propozycje przyszłych działań\n",
    "\n",
    "Ćwiczenie mogłoby zostać powtórzone z dodaniem nowych zbiorów danych, które mogły by rozszerzyć obecny o informacje takie jak liczba lokali w budynku/sąsiedztwie, czy dane demograficzne. \n",
    "\n",
    "Dodatkowo można by przeprowadzić głębszą analizę poprawności danych o infrastrukturze i zastosować dodatkowe kroki przetwarzania ze względu na to że występują spore różnice między tym jak poszczególni przedsiębiorcy się sprawozdają, alternatywnie przyjąć inne założenia co do wykorzystania danych o infrastrukturze, ponieważ tutaj dane podzieliliśmy tylko na 4 kategorie technologii podczas gdy technologii jest więcej. \n",
    "\n",
    "Poza tym można by spróbować określić skalę występowania jednego błędu w danych, który wcześniej nie został wspomniany. Błąd polega na tym, że dwóch dużych operatorów świadczy usługi w tych samych miejscach z czego jeden jest właścicielem, a drugi wynajmuje infrastrukturę od pierwszego. Dla tych samych adresów jeden operator potrafi oszacować maksymalną przepustowość na poziomie 80 mbit/s, a drugi 20 mbit/s (na tej samej infrastrukturze).\n",
    "\n",
    "## Podsumowanie\n",
    "\n",
    "W ramach tego projektu przygotowano zbiór danych w oparciu o otwarte źródła danych. W sumie pobrano i przetworzono kilkadziesiąt gigabajtów danych. Przetworzono je do wspólnego formatu, wyfiltrowano potrzebne informacje i połączono między sobą na podstawie atrybutów zarówno opisowych (adres) jak i przestrzennych (położenie). \n",
    "\n",
    "Niestety część źródeł, które chciałem wykorzystać była płatna jak np. dane z rejestru NOBC prowadzonego przez GUS, które wyceniono na 2500 zł brutto (na cele naukowe), a o część źródeł nie zdążyłem wystąpić jak np. zagregowane dane z rejestru PESEL.\n",
    "\n",
    "Jeżeli chodzi o modele uczenia maszynowego jak pokazały wyniki modele klasyfikacyjne nienajgorzej radzą sobie z tymi danymi osiągająć nawet ok. 80% dokładności na zbiorze testowym. Im wyższa klasa przepustowości tym lepsza była dokładność."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "---\n",
    "---\n",
    "Addendum:\n",
    "\n",
    "Wytrenowane modele można zapisać na dysku:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "from joblib import dump, load\n",
    "\n",
    "for k, v in models.items():\n",
    "    dump(v, k+'.joblib')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv('df.zip', index=False, sep=';')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}