naive_bayes.h

// SWAMI KARUPPASWAMI THUNNAI

#pragma once
#define DEBUG 1
#define NODEBUG 0
#define PI 3.141592653589793238462643383279502884L
#include <iostream>
#include <iomanip>
#include <string>
#include <sstream>
#include <fstream>
#include <vector>
#include <set>
#include <map>
#include "json.h"
#include <future>
#include <mutex>
#include <algorithm>
#include "dataioclass.h"
#include "itrainpredict.h"
using json = nlohmann::json;

/*
A class to store mean and variance. A datastructure to ease the math in gaussian_naive_bayes class.
*/
class mean_variance
{
private:
	unsigned long int column;
	double _mean;
	double variance;

public:
	mean_variance(unsigned long int column, double _mean, double variance) : column(column), _mean(_mean), variance(variance) {}

	unsigned long int get_column()
	{
		return column;
	}

	double get_mean()
	{
		return _mean;
	}

	double get_variance()
	{
		return variance;
	}
};

/*
Similar class of scikit-learn's GaussianNB
Written By: Visweswaran N on 2019-09-02
Edited By:  https://github.com/JUNZ1
*/
class gaussian_naive_bayes : public DataIOClass, public ITrainPredict
{

private:
	// Independent variable
	std::vector<std::vector<double>> X ={};
	// Dependent variable
	std::vector<unsigned long int> y = {};
	// Verbose to print debug messages
	unsigned short verbose;

	// Independent variable probabilities <features, probability>
	std::map<unsigned long int, std::map<double, double>> X_prob;

	// Dependent variable probabilities
	std::map<unsigned long int, double> y_prob;

	// Labels of y
	std::set<unsigned long int> labels;

	// label, column, mean and variance
	std::map<unsigned long int, std::vector<mean_variance>> mean_variance_map;

	

private:
	void print(std::string message);

	void calculate_y_probabilities();
	void calculate_x_probabilities();

public:
	/*
	Constructor to be used only when loading the existing model
	*/
	gaussian_naive_bayes(unsigned short verbose): verbose(verbose) {}
	/*
	Constructor to be used when creating a new model
	*/
	gaussian_naive_bayes(std::vector<std::vector<double>> X, std::vector<unsigned long int> y, unsigned short verbose): X(X), y(y), verbose(verbose){}

	gaussian_naive_bayes();

	gaussian_naive_bayes(const gaussian_naive_bayes& copyFromThis);

	gaussian_naive_bayes(gaussian_naive_bayes&& moveFromThis);
	virtual ~gaussian_naive_bayes(){};
	gaussian_naive_bayes& operator=(const gaussian_naive_bayes& copyFromThis);

	gaussian_naive_bayes& operator=(gaussian_naive_bayes&& moveFromThis);
	/*
	For fitting the model
	*/
	void fit();

	/*
	For predicting output
	*/
	std::map<unsigned long int, double> predict(std::vector<double> X_test);

	/*
	Used to save the model
	*/
	void save_model(std::string model_name);

	/*
	Used to load the saved model
	*/
	void load_model(std::string model_name);

public: //Overrited Interfaces
	void Train() override;
	std::vector<double> Predict(std::vector<double>) override;
private:
	std::mutex _trainingMutex;
	std::future<void> _trainFuture;
};