mining-offensive-speech.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.feature_extraction.text import TfidfTransformer\n",
    "from sklearn.metrics import confusion_matrix, f1_score\n",
    "from arff2pandas import a2p\n",
    "\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('OffComBR2.arff') as f, open('OffComBR3.arff') as g:\n",
    "    data1 = a2p.load(f)\n",
    "    data2 = a2p.load(f)\n",
    "    data = data1.append(data2, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.rename(columns = {'@@class@{no,yes}':'target','document@STRING':'text'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>target</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>yes</td>\n",
       "      <td>Votaram no PEZAO Agora tomem no CZAO</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>no</td>\n",
       "      <td>cuidado com a poupanca pessoal Lembram o que a...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>no</td>\n",
       "      <td>Sabe o que eu acho engracado os nossos governa...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>yes</td>\n",
       "      <td>os cariocas tem o que merecem um pessoal que s...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>no</td>\n",
       "      <td>Podiam retirar dos lucros dos bancos</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  target                                               text\n",
       "0    yes               Votaram no PEZAO Agora tomem no CZAO\n",
       "1     no  cuidado com a poupanca pessoal Lembram o que a...\n",
       "2     no  Sabe o que eu acho engracado os nossos governa...\n",
       "3    yes  os cariocas tem o que merecem um pessoal que s...\n",
       "4     no              Podiam retirar dos lucros dos bancos "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Quantos dados de cada tipo temos?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_data = data['target'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "no     831\n",
       "yes    419\n",
       "Name: target, dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x114da1da0>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.barplot(x=feature_data.values, y=feature_data.index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Quão longo são os textos? do tipo ofensivo e não ofensivo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "s1 = data[data['target'] == 'yes']['text'].str.len()\n",
    "s2 = data[data['target'] == 'no']['text'].str.len()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x114e34b38>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.distplot(s1, label='Offensive')\n",
    "sns.distplot(s2, label='Not Offensive')\n",
    "plt.title('Lenght Distribution')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "57.517899761336515 80.14320096269554\n"
     ]
    }
   ],
   "source": [
    "print(s1.mean(), s2.mean())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Conclusion:** Textos não ofensivos são geralmente mais longos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x114f79b70>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "d1 = data[data['target'] == 'yes']['text'].str.replace(r'\\d+', '').str.len()\n",
    "sns.distplot(d1, label='Offensive')\n",
    "d2 = data[data['target'] == 'no']['text'].str.replace(r'\\d+', '').str.len()\n",
    "sns.distplot(d2, label='Not Offensive')\n",
    "plt.title('Digits Distribution')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Offensive texts: 57.517899761336515\n",
      "Not Offensive texts: 80.14320096269554\n"
     ]
    }
   ],
   "source": [
    "print('Offensive texts: {n1}\\nNot Offensive texts: {n2}'.format(n1=d1.mean(), n2=d2.mean()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Obs:** Textos não ofensivos tem uma quantidade maior de digitos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1151819b0>"
      ]
     },
     "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": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "nd1 = data[data['target'] == 'yes']['text'].str.replace(r'\\w+', '').str.len()\n",
    "sns.distplot(nd1, label='Offensive')\n",
    "nd2 = data[data['target'] == 'no']['text'].str.replace(r'\\w+', '').str.len()\n",
    "sns.distplot(nd2, label='Not Offensive')\n",
    "plt.title('Non-Digits Distribution')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Offensive texts: 9.976133651551313\n",
      "Not Offensive texts: 14.110709987966306\n"
     ]
    }
   ],
   "source": [
    "print('Offensive texts: {n1}\\nNot Offensive texts: {n2}'.format(n1=nd1.mean(), n2=nd2.mean()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Obs:** Textos não ofensivos tem uma presença maior de não digitos."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Detalhes sobre a coluna target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"4\" halign=\"left\">text</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>unique</th>\n",
       "      <th>top</th>\n",
       "      <th>freq</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>target</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>no</th>\n",
       "      <td>831</td>\n",
       "      <td>828</td>\n",
       "      <td>PESQUISE E PARTICIPE Consulta Publica  PEC</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>yes</th>\n",
       "      <td>419</td>\n",
       "      <td>416</td>\n",
       "      <td>Pq o Morosita recebe ordens externas A ordem n...</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        text                                                               \n",
       "       count unique                                                top freq\n",
       "target                                                                     \n",
       "no       831    828        PESQUISE E PARTICIPE Consulta Publica  PEC     2\n",
       "yes      419    416  Pq o Morosita recebe ordens externas A ordem n...    2"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby('target').describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(data['text'], data['target'], random_state=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* **CountVectorizer**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary len: 3431\n",
      "Longest word: kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "\n",
    "vect = CountVectorizer().fit(X_train)\n",
    "print('Vocabulary len:', len(vect.get_feature_names()))\n",
    "print('Longest word:', max(vect.vocabulary_, key=len))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train_vectorized = vect.transform(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 1, 1, ..., 3, 1, 1])"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train_vectorized.data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "values = x_train_vectorized.data\n",
    "df = pd.DataFrame(data=values).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>10441</th>\n",
       "      <th>10442</th>\n",
       "      <th>10443</th>\n",
       "      <th>10444</th>\n",
       "      <th>10445</th>\n",
       "      <th>10446</th>\n",
       "      <th>10447</th>\n",
       "      <th>10448</th>\n",
       "      <th>10449</th>\n",
       "      <th>10450</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 10451 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   0      1      2      3      4      5      6      7      8      9      \\\n",
       "0      1      1      1      1      1      1      1      1      1      1   \n",
       "\n",
       "   ...    10441  10442  10443  10444  10445  10446  10447  10448  10449  10450  \n",
       "0  ...        1      1      1      1      1      1      1      3      1      1  \n",
       "\n",
       "[1 rows x 10451 columns]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=0.1, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.naive_bayes import MultinomialNB\n",
    "model = MultinomialNB(alpha=0.1)\n",
    "model.fit(x_train_vectorized, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['abaixo' 'abarrotadas' 'abc' ... 'zozoi' 'zuados' 'zuar']\n",
      "[   0 2000 2001 ... 2139 2679  856]\n"
     ]
    }
   ],
   "source": [
    "# get the feature names as numpy array\n",
    "feature_names = np.array(vect.get_feature_names())\n",
    "\n",
    "# Sort the coefficients from the model\n",
    "sorted_coef_index = model.coef_[0].argsort()\n",
    "\n",
    "print(feature_names)\n",
    "print(sorted_coef_index)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Smallest Coefs:\n",
      "['abaixo' 'men' 'mendigos' 'menguinho' 'meninos' 'mensagens' 'mente'\n",
      " 'mentir' 'melhores' 'mentira']\n",
      "\n",
      "Largest Coefs: \n",
      "['de' 'que' 'nao' 'um' 'do' 'vai' 'da' 'com' 'no' 'mais']\n"
     ]
    }
   ],
   "source": [
    "# Find the 10 smallest and 10 largest coefficients\n",
    "# The 10 largest coefficients are being indexed using [:-11:-1] \n",
    "# so the list returned is in order of largest to smallest\n",
    "\n",
    "print('Smallest Coefs:\\n{}\\n'.format(feature_names[sorted_coef_index[:10]]))\n",
    "print('Largest Coefs: \\n{}'.format(feature_names[sorted_coef_index[:-11:-1]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 74.12%\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "y_pred = model.predict(vect.transform(X_test))\n",
    "print('Accuracy: %.2f%%' % (accuracy_score(y_test, y_pred) * 100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* **Tfidf**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "vect = TfidfVectorizer(min_df=3).fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary len: 675\n",
      "Longest word: aposentadoria\n"
     ]
    }
   ],
   "source": [
    "print('Vocabulary len:', len(vect.get_feature_names()))\n",
    "print('Longest word:', max(vect.vocabulary_, key=len))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_vectorized = vect.transform(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.3137048 , 0.35805486, 0.39266908, ..., 0.2710915 , 0.13512801,\n",
       "       0.25406961])"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_vectorized.data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=0.1, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = MultinomialNB(alpha=0.1)\n",
    "model.fit(X_train_vectorized, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Smallest Coefs:\n",
      "['ministra' 'dizer' 'real' 'paga' 'reais' 'paises' 'duras' 'parabens'\n",
      " 'educacao' 'rajkazblanks']\n",
      "\n",
      "Largest Coefs: \n",
      "['de' 'que' 'vai' 'um' 'nao' 'do' 'idiota' 'da' 'seu' 'sua']\n"
     ]
    }
   ],
   "source": [
    "feature_names = np.array(vect.get_feature_names())\n",
    "\n",
    "sorted_coef_index = model.coef_[0].argsort()\n",
    "\n",
    "print('Smallest Coefs:\\n{}\\n'.format(feature_names[sorted_coef_index[:10]]))\n",
    "print('Largest Coefs: \\n{}'.format(feature_names[sorted_coef_index[:-11:-1]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 70.29%\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "y_pred = model.predict(vect.transform(X_test))\n",
    "print('Accuracy: %.2f%%' % (accuracy_score(y_test, y_pred) * 100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Feature Engineering**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def add_feature(X, feature_add):\n",
    "    \"\"\"\n",
    "    Returns sparse feature matrix with added feature.\n",
    "    feature_to_add can also be a list of features.\n",
    "    \"\"\"\n",
    "    from scipy.sparse import csr_matrix, hstack\n",
    "    return hstack([X, csr_matrix(feature_add).T], 'csr')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**First Model with**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, let's ignore terms that have a document frequency strictly lower than 3. Using this document-term matrix and an additional feature, the length of document (number of characters), we will test how our Tfidf performs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary len: 675\n",
      "Longest word: aposentadoria\n"
     ]
    }
   ],
   "source": [
    "vect = TfidfVectorizer(min_df=3).fit(X_train)\n",
    "print('Vocabulary len:', len(vect.get_feature_names()))\n",
    "print('Longest word:', max(vect.vocabulary_, key=len))\n",
    "\n",
    "X_train_vectorized = vect.transform(X_train)\n",
    "\n",
    "X_train_vectorized = add_feature(X_train_vectorized, X_train.str.len())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=0.1, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = MultinomialNB(alpha=0.1)\n",
    "model.fit(X_train_vectorized, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Smallest Coefs:\n",
      "['acabar', 'acaso', 'acham', 'acreditar', 'alcance', 'alguem', 'algum', 'alguma', 'alguns', 'am']\n",
      "\n",
      "Largest Coefs: \n",
      "['length_of_doc', 'de', 'que', 'vai', 'um', 'nao', 'do', 'idiota', 'da', 'seu']\n"
     ]
    }
   ],
   "source": [
    "index = np.array(vect.get_feature_names() + ['length_of_doc'])\n",
    "values  = model.coef_[0]\n",
    "features_series = pd.Series(data=values,index=index)\n",
    "\n",
    "print('Smallest Coefs:\\n{}\\n'.format(features_series.nsmallest(10).index.values.tolist()))\n",
    "print('Largest Coefs: \\n{}'.format(features_series.nlargest(10).index.values.tolist()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 70.29%\n"
     ]
    }
   ],
   "source": [
    "X_test_vectorized = vect.transform(X_test)\n",
    "X_test_vectorized = add_feature(X_test_vectorized, X_test.str.len())\n",
    "    \n",
    "y_pred = model.predict(X_test_vectorized)\n",
    "print('Accuracy: %.2f%%' % (accuracy_score(y_test, y_pred) * 100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Second model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tf-idf: \n",
    "* Ignorar termos com frequencia abaixo de 5\n",
    "* word n-grams n=1 to n=3 (unigrams, bigrams and trigrams)\n",
    "\n",
    "Features adicionais: \n",
    "\n",
    "* Tamanho do documento (quantidade de caracteres)\n",
    "* Número de digitos por documento"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary len: 413\n",
      "Longest word: em pontos corridos\n"
     ]
    }
   ],
   "source": [
    "vect = TfidfVectorizer(min_df=5, ngram_range=(1, 3)).fit(X_train)\n",
    "print('Vocabulary len:', len(vect.get_feature_names()))\n",
    "print('Longest word:', max(vect.vocabulary_, key=len))\n",
    "\n",
    "X_train_vectorized = vect.transform(X_train)\n",
    "\n",
    "X_train_vectorized = add_feature(X_train_vectorized, X_train.str.len())\n",
    "X_train_vectorized = add_feature(X_train_vectorized, X_train.str.replace(r'\\D+', '').str.len())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=0.1, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = MultinomialNB(alpha=0.1)\n",
    "model.fit(X_train_vectorized, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Smallest Coefs:\n",
      "['acabar', 'acabar com', 'alguem', 'alguma', 'alguns', 'anos de', 'aos', 'aumento', 'colocar', 'com os']\n",
      "\n",
      "Largest Coefs: \n",
      "['length_of_doc', 'de', 'que', 'vai', 'um', 'do', 'nao', 'idiota', 'da', 'seu']\n"
     ]
    }
   ],
   "source": [
    "index = np.array(vect.get_feature_names() + ['length_of_doc', 'digit_count'])\n",
    "values  = model.coef_[0]\n",
    "features_series = pd.Series(data=values,index=index)\n",
    "\n",
    "print('Smallest Coefs:\\n{}\\n'.format(features_series.nsmallest(10).index.values.tolist()))\n",
    "print('Largest Coefs: \\n{}'.format(features_series.nlargest(10).index.values.tolist()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 69.33%\n"
     ]
    }
   ],
   "source": [
    "X_test_vectorized = vect.transform(X_test)\n",
    "X_test_vectorized = add_feature(X_test_vectorized, X_test.str.len())\n",
    "X_test_vectorized = add_feature(X_test_vectorized, X_test.str.replace(r'\\D+', '').str.len())\n",
    "    \n",
    "y_pred = model.predict(X_test_vectorized)\n",
    "print('Accuracy: %.2f%%' % (accuracy_score(y_test, y_pred) * 100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Third model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Count Vectorizer  \n",
    "* Termos com frequência abaixo de 5  \n",
    "* N-grams de n=2 a n=5  \n",
    "\n",
    "To tell Count Vectorizer to use character n-grams we pass in analyzer='char_wb' which creates character n-grams only from text inside word boundaries. This should make the model more robust to spelling mistakes.\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "Features Adicionais:\n",
    "\n",
    "* Tamanho do documento (número de caracteres)  \n",
    "* Número de digítos por caracter\n",
    "* Número de digitos que não sejam palavras (Letras, digitos ou underscore.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary len: 20574\n",
      "Longest word:  esse\n"
     ]
    }
   ],
   "source": [
    "vect = CountVectorizer(min_df=1, ngram_range=(2, 5), analyzer='char_wb').fit(X_train)\n",
    "print('Vocabulary len:', len(vect.get_feature_names()))\n",
    "print('Longest word:', max(vect.vocabulary_, key=len))\n",
    "\n",
    "X_train_vectorized = vect.transform(X_train)\n",
    "\n",
    "X_train_vectorized = add_feature(X_train_vectorized, X_train.str.len())\n",
    "X_train_vectorized = add_feature(X_train_vectorized, X_train.str.replace(r'\\D+', '').str.len())\n",
    "X_train_vectorized = add_feature(X_train_vectorized, X_train.str.replace(r'\\w+', '').str.len())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=0.1, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = MultinomialNB(alpha=0.1)\n",
    "model.fit(X_train_vectorized, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Smallest Coefs:\n",
      "[' aba', ' abai', ' abar', ' abc', ' abc ', ' abs', ' absu', ' acao', ' acas', ' acel']\n",
      "\n",
      "Largest Coefs: \n",
      "['length_of_doc', 'non_word_char_count', 'a ', 'o ', 'e ', 'kk', 'kkk', 'kkkk', ' e', ' d']\n"
     ]
    }
   ],
   "source": [
    "index = np.array(vect.get_feature_names() + ['length_of_doc', 'digit_count', 'non_word_char_count'])\n",
    "values = model.coef_[0]\n",
    "features_series = pd.Series(data=values,index=index)\n",
    "\n",
    "print('Smallest Coefs:\\n{}\\n'.format(features_series.nsmallest(10).index.values.tolist()))\n",
    "print('Largest Coefs: \\n{}'.format(features_series.nlargest(10).index.values.tolist()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 75.40%\n"
     ]
    }
   ],
   "source": [
    "X_test_vectorized = vect.transform(X_test)\n",
    "X_test_vectorized = add_feature(X_test_vectorized, X_test.str.len())\n",
    "X_test_vectorized = add_feature(X_test_vectorized, X_test.str.replace(r'\\D+', '').str.len())\n",
    "X_test_vectorized = add_feature(X_test_vectorized, X_test.str.replace(r'\\w+', '').str.len())\n",
    "    \n",
    "y_pred = model.predict(X_test_vectorized)\n",
    "print('Accuracy: %.2f%%' % (accuracy_score(y_test, y_pred) * 100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}