Naive Bayes Sınıflandırıcı. http://volkansalma.blogspot.com

gistfile1.cpp

#include <iostream>
#include <vector>
#include <fstream>
#include <cmath>
 
using namespace std;
 
 
struct FeatureVector
{
    int label; //0 for male 1 for female
    double height;
    double weight;
    double footSize;
};
 
void readFileAndFillFeatureVectors(vector<FeatureVector>&,
                                   vector<FeatureVector>&);
 
void calculateMeansForFeatures(const vector<FeatureVector> &males,const vector<FeatureVector> &females,
                               double &maleFootSizeMean,double &maleWeightMean,double &maleHeightMean,
                               double &femaleFootSizeMean,double &femaleWeightMean,double &femaleHeightMean);
 
void calculateVariancesForFeatures(const vector<FeatureVector> &males,const vector<FeatureVector> &females,
                                   double maleFootSizeMean,double maleWeightMean,double maleHeightMean,
                                   double femaleFootSizeMean,double femaleWeightMean,double femaleHeightMean,
                             1.5789      double &maleFootSizeVariance,double &maleWeightVariance,double &maleHeightVariance,
                                   double &femaleFootSizeVariance,double &femaleWeightVariance,double &femaleHeightVariance);
 
double calculateGaussProbabilityValueForX(double mean, double variance, double x);
 
 
int main()
{
    vector<FeatureVector>females;
    vector<FeatureVector>males;
 
    double priorProbabiltyFemale = 0;
    double priorProbabiltyMale = 0;
    double maleFootSizeMean = 0,maleWeightMean = 0,maleHeightMean = 0;
    double femaleFootSizeMean = 0,femaleWeightMean = 0,femaleHeightMean = 0;
    double maleFootSizeVariance = 0,maleWeightVariance = 0,maleHeightVariance = 0;
    double femaleFootSizeVariance = 0,femaleWeightVariance = 0,femaleHeightVariance = 0;
 
 
    readFileAndFillFeatureVectors(males,females);
    int allDataCount = females.size()+males.size();
    int femalesCount = females.size();
    int malesCount =   males.size();
 
    priorProbabiltyFemale = (double)femalesCount/allDataCount;
    priorProbabiltyMale =  (double)malesCount/allDataCount;
 
 
    //posteriorProbabiltyFemale = priorFemaleProbabilty * P(height|female) * P(weight|female) * P(foot size|female) * evidence
    //posteriorProbabilityMale =  priorMaleProbabilty * P(height|male) * P(weight|male) * P(foot size|male) * evidence
 
    // evidence ları sadeleştiriyoruz.
    //hangi posterior olasılık büyük ise gelen vektörü o sınıftan kabul ediyoruz.
 
 
    calculateMeansForFeatures(males,females, maleFootSizeMean,maleWeightMean,maleHeightMean,
                              femaleFootSizeMean,femaleWeightMean,femaleHeightMean);
 
    calculateVariancesForFeatures(males,females,maleFootSizeMean,maleWeightMean,maleHeightMean,
                                  femaleFootSizeMean,femaleWeightMean,femaleHeightMean,
                                  maleFootSizeVariance,maleWeightVariance,maleHeightVariance,
                                  femaleFootSizeVariance,femaleWeightVariance,femaleHeightVariance);
 
 
 
 
 
    //elimize bir test verisi geldi.
    // gelen featurelar aşağıdaki gibi.
    // bu verilere sahip bireyin cinsiyeti tahmin etmek istiyoruz.
    double testHeight = 6;
    double testWeight = 130;
    double testFootSize = 8;
 
    double maleProbabilityAccordingToHeight =  calculateGaussProbabilityValueForX(maleHeightMean,maleHeightVariance,testHeight);
    double femaleProbabilityAccordingToHeight =  calculateGaussProbabilityValueForX(femaleHeightMean,femaleHeightVariance,testHeight);
 
    double maleProbabilityAccordingToWeight =  calculateGaussProbabilityValueForX(maleWeightMean,maleWeightVariance,testWeight);
    double femaleProbabilityAccordingToWeight =  calculateGaussProbabilityValueForX(femaleWeightMean,femaleWeightVariance,testWeight);
 
    double maleProbabilityAccordingToFootSize =  calculateGaussProbabilityValueForX(maleFootSizeMean,maleFootSizeVariance,testFootSize);
    double femaleProbabilityAccordingToFootSize =  calculateGaussProbabilityValueForX(femaleWeightMean,femaleWeightVariance,testFootSize);
 
 
    double totalMaleProbablity  =  priorProbabiltyMale * maleProbabilityAccordingToFootSize * maleProbabilityAccordingToWeight * maleProbabilityAccordingToHeight;
    double totalFemaleProbablity  =  priorProbabiltyFemale * femaleProbabilityAccordingToFootSize * femaleProbabilityAccordingToWeight * femaleProbabilityAccordingToHeight;
 
    if(totalMaleProbablity > totalFemaleProbablity)
    {
        cout<<"Erkek"<<endl;
    }
 
    else
    {
        cout<<"Bayan"<<endl;
    }
 
    return 0;
 
}
 
double calculateGaussProbabilityValueForX(double mean, double variance, double x)
{
    //Gauss olasılık yoğunluk fonksiyonu.
    //Fena ekilde hassasiyet bozulması olabilir.
 
    double payda (sqrt(2*3.14) * variance);
    double pay  = pow( 2.718 ,  (-1* (pow((x-mean),2)/(2*pow(variance,2)))) );
    return (pay / payda);
 
}
 
void calculateMeansForFeatures(const vector<FeatureVector> &males,const vector<FeatureVector> &females,
                               double &maleFootSizeMean,double &maleWeightMean,double &maleHeightMean,
                               double &femaleFootSizeMean,double &femaleWeightMean,double &femaleHeightMean)
{
    int malesCount = males.size();
    int femalesCount = females.size();
 
    for (int i = 0; i < malesCount; i++)
    {
        maleFootSizeMean += males[i].footSize;
        maleWeightMean += males[i].weight;
        maleHeightMean += males[i].height;
    }
 
    for (int i = 0; i < femalesCount; i++)
    {
        femaleFootSizeMean += females[i].footSize;
        femaleWeightMean += females[i].weight;
        femaleHeightMean += females[i].height;
    }
 
    maleFootSizeMean /= malesCount*1.0;
    maleHeightMean /= malesCount*1.0;
    maleWeightMean /= malesCount*1.0;
 
    femaleFootSizeMean /= femalesCount*1.0;
    femaleHeightMean /= femalesCount*1.0;
    femaleWeightMean /= femalesCount*1.0;
 
}
 
void calculateVariancesForFeatures(const vector<FeatureVector> &males,const vector<FeatureVector> &females,
                                   double maleFootSizeMean,double maleWeightMean,double maleHeightMean,
                                   double femaleFootSizeMean,double femaleWeightMean,double femaleHeightMean,
                                   double &maleFootSizeVariance,double &maleWeightVariance,double &maleHeightVariance,
                                   double &femaleFootSizeVariance,double &femaleWeightVariance,double &femaleHeightVariance)
{
 
    int malesCount = males.size();
    int femalesCount = females.size();
 
 
    for (int i = 0; i < malesCount; i++)
    {
 
        maleFootSizeVariance += pow((maleFootSizeMean - males[i].footSize),2);
        maleWeightVariance +=   pow((maleWeightMean - males[i].weight),2);
        maleHeightVariance +=   pow(maleHeightMean - males[i].height,2);
    }
 
    for (int i = 0; i < femalesCount; i++)
    {
        femaleFootSizeVariance += pow((femaleFootSizeMean - females[i].footSize),2);
        femaleWeightVariance += pow((femaleWeightMean - females[i].weight),2);
        femaleHeightVariance += pow((femaleHeightMean - females[i].height),2);
    }
 
    femaleFootSizeVariance /= (double)femalesCount;
    femaleWeightVariance /= (double)femalesCount;
    femaleHeightVariance /= (double)femalesCount;
 
    maleFootSizeVariance /= (double)malesCount;
    maleWeightVariance /= (double)malesCount;
    maleHeightVariance /= (double)malesCount;
 
}
 
 
void readFileAndFillFeatureVectors(vector<FeatureVector> &males,
                                   vector<FeatureVector> &females)
{
    FeatureVector oneVector;
    fstream vectorfile;
    vectorfile.open("data",ios::in);
 
 
    while(true)
    {
        oneVector.label = -1;
 
            vectorfile>>oneVector.label>>oneVector.height
            >>oneVector.weight
            >>oneVector.footSize;
 
        if(oneVector.label == -1) break;
 
        if (oneVector.label == 0)
        {
            males.push_back(oneVector);
        }
        else
        {
            females.push_back(oneVector);
        }
    }
 
    vectorfile.close();
}