From 4e18fda54aa0cb811969a24e33239cc44292d819 Mon Sep 17 00:00:00 2001
From: Vishisht Priyadarshi <vishishtpriyadarshi867@gmail.com>
Date: Tue, 14 Dec 2021 08:34:04 +0000
Subject: [PATCH] Improve package

---
 .gitignore                     |  2 +-
 cobraclassifier/cobra_boost.py | 66 ++++++++++++++++++++++++++++++++++
 setup.py                       |  6 ++--
 3 files changed, 70 insertions(+), 4 deletions(-)
 create mode 100644 cobraclassifier/cobra_boost.py

diff --git a/.gitignore b/.gitignore
index 4477699..a1f5910 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,5 +1,5 @@
 .venv/
-__pycache__/
+**/__pycache__/
 .vscode
 build/
 dist/
\ No newline at end of file
diff --git a/cobraclassifier/cobra_boost.py b/cobraclassifier/cobra_boost.py
new file mode 100644
index 0000000..9abd816
--- /dev/null
+++ b/cobraclassifier/cobra_boost.py
@@ -0,0 +1,66 @@
+import numpy as np
+from math import log
+from itertools import compress
+
+from cobraclassifier import classifier_cobra as cobra
+from cobraclassifier import edited_knn, near_miss_v1, near_miss_v2, near_miss_v3, tomek_link, condensed_knn, knn_und
+
+
+class CobraBoost:
+    def __init__(self, X, y, machines, undersampling_method):
+        self.X = X
+        self.y = y
+
+        self.model = cobra(machines = machines)
+        self.majority_class_label = int(sum(y) > 0.5 * len(y))
+        self.undersampling_method = undersampling_method
+
+        self.weight_update = 0
+        self.init_w = 1.0 / len(self.X)
+        self.weight = np.full(len(self.X), self.init_w)
+        
+
+    def learn_parameters(self, iterations):
+        verdict = self.undersampling_method.undersample(self.X, self.y, self.majority_class_label)
+        X_undersampled, y_undersampled = self.X[verdict, :], self.y[verdict]
+    
+        for t in range(iterations):
+            print("[Testing]: Executing the iteration - {} of CobraBoost".format(t + 1))
+            
+            self.model.fit(X_undersampled, y_undersampled, sample_weight = self.weight[verdict])
+        
+            flag = self.y != self.model.predict(self.X)
+            loss = sum(list(compress(self.weight, flag)))
+            
+            alpha = loss / (1 - loss)  
+
+            if alpha <= 0:
+                alpha = 0.0000001
+            else:
+                try:
+                    alpha_hat = 0.5 * (np.log(1 - loss) - np.log(loss))
+                except:
+                    alpha_hat = 0
+                    
+                self.weight = self.weight * np.exp(-alpha_hat * self.y * self.model.predict(self.X))
+                self.weight = self.weight / self.weight.sum()
+
+            self.weight_update = alpha
+            
+
+    def predict(self, test_data):
+        n = len(test_data)
+        predicted_labels = np.zeros(n)
+
+        for i in range(n):
+            positive_score, negative_score = 0, 0
+            
+            if self.model.predict(test_data[i].reshape(1, -1)) == 1:
+                positive_score += log(1/self.weight_update)
+            else:
+                negative_score += log(1/self.weight_update)
+            
+            if negative_score <= positive_score:
+                predicted_labels[i] = 1
+
+        return predicted_labels
\ No newline at end of file
diff --git a/setup.py b/setup.py
index 2a04957..99d0a2d 100644
--- a/setup.py
+++ b/setup.py
@@ -2,13 +2,13 @@
 setup(
   name = 'cobraclassifier',
   packages = ['cobraclassifier'],
-  version = '1.3',
-  license='MIT', 
+  version = '1.4',
+  license = 'MIT', 
   description = 'COBRA for classification tasks (on Imbalanced Data)',
   author = ['Dr. Arabin Kumar Dey', 'Vishisht Priyadarshi', 'Aadi Gupta', 'Tejus Singla', 'Shashank Goyal'],
   author_email = 'vishishtpriyadarshi867@gmail.com',
   url = 'https://github.com/vishishtpriyadarshi/MA691-COBRA-6',
-  download_url = 'https://github.com/vishishtpriyadarshi/MA691-COBRA-6/archive/refs/tags/v1.1.tar.gz',
+  download_url = 'https://github.com/vishishtpriyadarshi/MA691-COBRA-6/archive/refs/tags/v1.4.tar.gz',
   keywords = ['Classification', 'Imbalanced Data', 'Machine Learning'],
   install_requires=[
           'numpy',