From b33975437c080278d1f0a48e9cb241dc27d99ed6 Mon Sep 17 00:00:00 2001
From: SARTHAK <shibugarg0303@gmail.com>
Date: Mon, 6 Feb 2023 00:34:47 +0530
Subject: [PATCH] Added multinomial nb

---
 CMakeLists.txt                                |   5 +-
 examples/neighbors/bernoulli_nb.cpp           |   3 +-
 examples/neighbors/multinomial_nb.cpp         |  24 +++++
 src/slowmokit.hpp                             |   1 +
 .../neighbors/bernoulli_nb/bernoulli_nb.cpp   |   8 +-
 .../neighbors/bernoulli_nb/bernoulli_nb.hpp   |   5 +-
 .../methods/neighbors/multinomial_nb.hpp      |  13 +++
 .../multinomial_nb/multinomial_nb.cpp         | 100 ++++++++++++++++++
 .../multinomial_nb/multinomial_nb.hpp         |  25 +++++
 9 files changed, 180 insertions(+), 4 deletions(-)
 create mode 100644 examples/neighbors/multinomial_nb.cpp
 create mode 100644 src/slowmokit/methods/neighbors/multinomial_nb.hpp
 create mode 100644 src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.cpp
 create mode 100644 src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.hpp

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 4b03d30..7781392 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -39,4 +39,7 @@ add_library(slowmokit
         src/slowmokit/methods/preprocessing/standardization.cpp
         src/slowmokit/methods/neighbors/bernoulli_nb.hpp
         src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.hpp
-        src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.cpp)
\ No newline at end of file
+        src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.cpp
+        src/slowmokit/methods/neighbors/multinomial_nb.hpp
+        src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.hpp
+        src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.cpp)
\ No newline at end of file
diff --git a/examples/neighbors/bernoulli_nb.cpp b/examples/neighbors/bernoulli_nb.cpp
index 0d00148..69caa4d 100644
--- a/examples/neighbors/bernoulli_nb.cpp
+++ b/examples/neighbors/bernoulli_nb.cpp
@@ -17,6 +17,7 @@
 //     std::vector<int> yTrain{0,0,1,1,0,0,1,1,1,0};
 //     std::vector<int> xTest{1,0,0};
 //     BernoulliNB<int> bernoulli;
-//     std::cout<<bernoulli.predict(xTrain,yTrain,xTest);
+//     bernoulli.fit(xTrain,yTrain);
+//     std::cout<<bernoulli.fit_predict(xTest);
 //     return 0;
 // }
\ No newline at end of file
diff --git a/examples/neighbors/multinomial_nb.cpp b/examples/neighbors/multinomial_nb.cpp
new file mode 100644
index 0000000..e447488
--- /dev/null
+++ b/examples/neighbors/multinomial_nb.cpp
@@ -0,0 +1,24 @@
+// #include "../../src/slowmokit/methods/neighbors/multinomial_nb.hpp"
+// #include "../../src/slowmokit/core.hpp"
+
+// signed main(){
+//     std::vector<std::vector<int>> xTrain{
+//         {0,0,0},
+//         {1,0,0},
+//         {1,0,1},
+//         {1,0,0},
+//         {0,1,1},
+//         {1,0,0},
+//         {1,1,0},
+//         {0,1,1},
+//         {1,0,1},
+//         {1,1,1}
+//     };
+//     std::vector<int> yTrain{2,2,2,2,1,2,2,1,2,0};
+//     std::vector<int> xTest{0,0,0};
+//     std::vector<int> classes{0,1,2};
+//     MultinomialNB<int> multinomial;
+//     multinomial.fit(xTrain,yTrain);
+//     std::cout<<multinomial.predict(xTest,classes);
+//     return 0;
+// }
\ No newline at end of file
diff --git a/src/slowmokit.hpp b/src/slowmokit.hpp
index 1d929d9..f56d33e 100644
--- a/src/slowmokit.hpp
+++ b/src/slowmokit.hpp
@@ -18,5 +18,6 @@
 #include "slowmokit/methods/cluster/kMeans.hpp"
 #include "slowmokit/methods/linear_model/linear_regression.hpp"
 #include "slowmokit/methods/neighbors/bernoulli_nb.hpp"
+#include "slowmokit/methods/neighbors/multinomial_nb.hpp"
 
 #endif // SLOWMOKIT_HPP
diff --git a/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.cpp b/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.cpp
index e2f793f..98561cc 100644
--- a/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.cpp
+++ b/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.cpp
@@ -36,7 +36,13 @@ template<class T>
     }
 
 template<class T>
-    int fit(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain,std::vector<T> xTest){
+    void fit(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain){
+        this->xTrain = xTrain;
+        this->yTrain = yTrain;
+    }
+
+template<class T>
+    int fit_predict(std::vector<T> xTest){
         int nFeatures = xTrain[0].size();
 
         std::vector<double> postProbs;
diff --git a/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.hpp b/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.hpp
index 0df9b4c..1ffe1bc 100644
--- a/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.hpp
+++ b/src/slowmokit/methods/neighbors/bernoulli_nb/bernoulli_nb.hpp
@@ -12,12 +12,15 @@
 template<class T>
 class BernoulliNB{  
 private:
+    std::vector<std::vector<int>> xTrain;
+    std::vector<int> yTrain;
     double prior_prob(std::vector<T> yTrain,int label);
     double conditional_prob(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain,int featureCol,int featureVal,int label);
 
 public:
     // returns predicted value for each test according to bernoulli naive bayes model
-    int fit(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain,std::vector<T> xTest);
+    void fit(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain)
+    int fit_predict(std::vector<T> xTest);
 };
 
 #endif
\ No newline at end of file
diff --git a/src/slowmokit/methods/neighbors/multinomial_nb.hpp b/src/slowmokit/methods/neighbors/multinomial_nb.hpp
new file mode 100644
index 0000000..f0f6fc0
--- /dev/null
+++ b/src/slowmokit/methods/neighbors/multinomial_nb.hpp
@@ -0,0 +1,13 @@
+/**
+ * @file methods/neighbors/multinomial_nb.hpp
+ *
+ * Easy include for Multinomial Nb algorithm
+ */
+
+
+#ifndef SLOWMOKIT_MULTINOMIAL_NB_HPP
+#define SLOWMOKIT_MULTINOMIAL_NB_HPP
+
+#include "multinomial_nb/multinomial_nb.hpp"
+
+#endif //SLOWMOKIT_MULTINOMIAL_NB_HPP
\ No newline at end of file
diff --git a/src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.cpp b/src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.cpp
new file mode 100644
index 0000000..822e985
--- /dev/null
+++ b/src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.cpp
@@ -0,0 +1,100 @@
+/**
+ * @file methods/neighbors/multinomial_nb/multinomial_nb.cpp
+ *
+ * Implementation of the Multinomial Naive Bayes main program
+ */
+#include "multinomial_nb.hpp"
+
+template<class T>
+    double priorProb(std::vector<T> yTrain,int label){  // Prior-Probability P(y)
+        int sum=0;
+        for(int i=0;i<yTrain.size();i++){
+            if(yTrain[i]==label){
+                sum += 1;
+            }
+        }
+        return sum/double(yTrain.size());
+    }
+
+template<class T>
+    double conditionalProb(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain,int featureCol,int featureVal,int label){  // Conditional-Probabilty P(x=f1 / y=class)
+        int denominator=0,numerator=0;
+        std::set<int> varies_values;
+        for(int i=0;i<yTrain.size();i++){
+            if(yTrain[i]==label){  // Check if label of feature is same as output looking for
+                denominator++;      // All values with output_class we are looking for
+                varies_values.insert(xTrain[i][featureCol]);
+                if(xTrain[i][featureCol]==featureVal){
+                    numerator++;       // numerator increase only when that feature consists of value same as value in testing feature
+                }
+            }
+        }
+        return (numerator+1)/double(denominator+varies_values.size());  // as it is multinomial nb so to avoid overfitting
+        // numerator+1 -> because if word do not exist in training but exist in testing
+        // denominator + number_of_different_values_in_feature
+    }
+
+template<class T>
+    void fit(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain,std::vector<T> classes){
+        this->xTrain = xTrain;
+        this->yTrain = yTrain;
+        this->classes = classes;
+    }
+
+template<class T>
+    int fit_predict(std::vector<T> xTest){
+        int n_features = xTrain[0].size();
+
+        std::vector<double> postProbs;
+
+        std::vector<int> ::iterator label;  // different ouput labels
+        for(label=classes.begin();label!=classes.end();label++){   // iterating over all labels (P(y=y1,y2,y3....))
+            double likelihood=1.0;  
+            for(int i=0;i<n_features;i++){
+                double cond = conditionalProb(xTrain,yTrain,i,xTest[i],*label);
+                likelihood += log10(cond);   // Log10 of computed conditionl probabilty => Log10(P(x=x1 / y=label))
+            }
+
+            double prior = priorProb(yTrain,*label);
+            double post = prior*likelihood;
+            postProbs.push_back(post);
+        }
+        double sumpropProbs = 0.0;
+        int max=0;
+        for(int i=0;i<postProbs.size();i++){  // Iterating over all labels computed values to check if any negative value exists, if so then direclty values are compared else we will compute value/sum
+            if(postProbs[i]<0){
+                sumpropProbs = 1;
+                break;
+            }
+            sumpropProbs += postProbs[i];
+        }
+        for(int i=0;i<postProbs.size();i++){
+            postProbs[i] /= sumpropProbs;
+            if(postProbs[i]>postProbs[max]){
+                max = i;
+            }
+        }
+        return max;
+    };
+
+signed main(){
+    std::vector<std::vector<int>> x_train{
+        {0,0,0},
+        {1,0,0},
+        {1,0,1},
+        {1,0,0},
+        {0,1,1},
+        {1,0,0},
+        {1,1,0},
+        {0,1,1},
+        {1,0,1},
+        {1,1,1}
+    };
+    std::vector<int> y_train{2,2,2,2,1,2,2,1,2,0};
+    std::vector<int> x_test{0,0,0};
+    std::vector<int> classes{0,1,2};
+    MultinomialNB<int> multinomial;
+    multinomial.fit(x_train,y_train,classes);
+    std::cout<<multinomial.fit_predict(x_test);
+    return 0;
+}
\ No newline at end of file
diff --git a/src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.hpp b/src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.hpp
new file mode 100644
index 0000000..34b03ba
--- /dev/null
+++ b/src/slowmokit/methods/neighbors/multinomial_nb/multinomial_nb.hpp
@@ -0,0 +1,25 @@
+/**
+ * @file methods/neighbors/multinomial_nb/multinomial_nb.hpp
+ *
+ * The header file including the Multinomial Naive Bayes algorithm
+ */
+
+#ifndef SLOWMOKIT_MULTINOMIAL_NB_HPP
+#define SLOWMOKIT_MULTINOMIAL_NB_HPP
+
+#include "../../../core.hpp"
+
+template<class T>
+class MultinomialNB{
+private:
+    std::vector<std::vector<T>> xTrain;
+    std::vector<T> yTrain;
+    std::vector<T> classes;
+    double priorProb(std::vector<T> yTrain,int label);
+    double conditionalProb(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain,int featureCol,int featureVal,int label);
+public:
+    void fit(std::vector<std::vector<T>> xTrain,std::vector<T> yTrain,std::vector<T> classes);
+    int fit_predict(std::vector<T> xTest);
+};
+
+#endif
\ No newline at end of file