From 742f3883cfd2367a766109999f58a814905be320 Mon Sep 17 00:00:00 2001
From: yacth <71322097+yacth@users.noreply.github.com>
Date: Wed, 16 Sep 2020 12:43:56 +0200
Subject: [PATCH] Update naive_bayes.py

---
 .../supervised_learning/naive_bayes.py        | 33 +++++++++++++++++++
 1 file changed, 33 insertions(+)

diff --git a/mlfromscratch/supervised_learning/naive_bayes.py b/mlfromscratch/supervised_learning/naive_bayes.py
index 9cbc4e08..790488d2 100644
--- a/mlfromscratch/supervised_learning/naive_bayes.py
+++ b/mlfromscratch/supervised_learning/naive_bayes.py
@@ -68,3 +68,36 @@ def predict(self, X):
         """ Predict the class labels of the samples in X """
         y_pred = [self._classify(sample) for sample in X]
         return y_pred
+    
+class BernoulliNaiveBayes:
+    def __init__(self,X,y):
+        self.X = X
+        self.y = y
+        self.N, self.D = X.shape
+        self.classes = np.unique(y)
+        self.C = len(self.classes)
+        self.prior = np.zeros(self.C,dtype=np.float64)
+        self.likelihood = np.zeros((self.C,self.D),dtype=np.float64)
+        self.N1 = sum(self.y)
+
+    def fit(self,X,y):        # Does the fit by calculating the prior and the likelihood of the training data
+        for c in self.classes:
+            X_c = X[c == y]
+            X_c_sum = np.sum(X_c, axis=0)
+            self.prior[c] = X_c.shape[0] / float(self.N)
+            for i in range(len(self.likelihood[0])):
+                self.likelihood[c,i] = X_c_sum[i]/len(X_c)
+
+    def predict(self,X_test): #Gives the predicted labels for dataset
+        y_pred = [self._predict(x) for x in X_test]
+        return y_pred
+
+    def _predict(self, x):
+        posteriors = []
+        for i, c in enumerate(self.classes):
+            log_prior = np.log1p(self.prior[i])
+            log_likelihood = np.sum(np.log1p(self.likelihood[i]*x)) + np.sum(np.log1p(self.likelihood[i]*(1-x)))
+            posterior = log_prior + log_likelihood
+            posteriors.append(posterior)
+        return self.classes[np.argmax(posteriors)]     
+