Skip to content

Add files via upload #3

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
stasysp wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Add files via upload #3

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
174 changes: 165 additions & 9 deletions classifier.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,7 @@
#include "classifier.hpp"
#include <cassert>
#include <map>


Classifier::Classifier(){
mathModel = 0;
Expand All @@ -11,9 +14,29 @@ Classifier::Classifier(){
ExH[0]=0; ExH[1]=0;
DxH[0]=0; DxH[1]=0;
H = 1e-7;
k = 0;
}
Classifier::~Classifier(){}

Classifier::Classifier(const Classifier &A){
mathModel = A.mathModel;
H=A.H;
for (int i=0; i<2; i++){
p[i]=A.p[i];
counts[i]=A.counts[i];
ExW[i]=A.ExW[i];
DxW[i]=A.DxW[i];
ExH[i]=A.ExH[i];
DxH[i]=A.DxH[i];
for (int j=0; j<250; j++){
pWeight[i][j]=A.pWeight[i][j];
pHeight[i][j]=A.pHeight[i][j];
}
}
for (int i=0; i<A.train_set.size(); i++)
train_set.push_back(A.train_set[i]);
}

void Classifier:: readTrainFile(ifstream &F){
while (!F.eof()){
d tmp;
Expand All @@ -25,7 +48,7 @@ void Classifier:: readTrainFile(ifstream &F){
vector<pair<double, double>> Classifier::readTestFile(ifstream &F){
vector<pair<double, double>> r;
while (!F.eof()){
int x, y;
double x, y;
F>>x>>y;
r.push_back(make_pair(x, y));
}
Expand All @@ -35,7 +58,7 @@ vector<pair<double, double>> Classifier::readTestFile(ifstream &F){

int Classifier::findNext(int i, double *arr){
for (; i<250; i++){
if (arr[i] > 10e-8) return i;
if (arr[i] > 1e-9) return i;
}
return 0;
}
Expand All @@ -49,6 +72,21 @@ double kernel(double _h, int index, int label, vector<d>data){
}
return f;
}
double kernel1(double _h, int index, int label, vector<d>data){
// sqrt((x-xi)^2+(y-yi)^2)
//K(x)=----------------------- if sqrt((x-xi)^2+(y-yi)^2) < h
// pi*h^2
//
const double pi = 3.14159265;
double f=0.0;
for (int i=0; i<data.size(); i++){
if (i!=index && data[i].label==label){
double tmp = sqrt(pow(data[index].weight-data[i].weight, 2)+pow(data[index].height-data[i].height, 2));
if (tmp<_h) f+= tmp/(pi*pow(_h, 2));
}
}
return f;
}
double Classifier::probability(double h, double w, int label){
// P(x|yi)=P(x)*P(yi|x)/P(yi)
// события yi независимые
Expand All @@ -65,7 +103,8 @@ double Classifier::probability(double h, double w, int label){
double pW = pow(exp, (-pow(w-ExW[label], 2)/(2*DxW[label])))/(sqrt(2*pi*DxW[label]));
double pH = pow(exp, (-pow(h-ExH[label], 2)/(2*DxH[label])))/(sqrt(2*pi*DxH[label]));
return pW*pH*p[label];
} else{
}
else if (mathModel == 3){
double f=0.0;
const double exp = 2.718;
for (int i=0; i<train_set.size(); i++){
Expand All @@ -74,21 +113,43 @@ double Classifier::probability(double h, double w, int label){
}
}
return p[label]*f/(counts[label]*H);
} else{ //mathModel == 5
const double pi = 3.14159265;
double f=0.0;
for (int i=0; i<train_set.size(); i++){
if (train_set[i].label==label){
double tmp = sqrt(pow(h-train_set[i].height, 2)+pow(w-train_set[i].weight, 2));
if (tmp<H) f+= tmp/(pi*pow(H, 2));
}
}
return p[label]*f/(counts[label]*(pi*pow(H, 2)));
}
}
int findAns(int _k, vector<pair<double, int>> data){
int count = 0;
for (int i=0; i<_k; i++)
if (data[i].second == 1) count++;
if (count > _k-count) return 1;
return 0;
}

Classifier barChart(Classifier A){
A.mathModel = 1;
int weight[2][250], heihgt[2][250];
int weight[2][250], height[2][250];
const double eps = 1e-8;
for (int i=0; i<2; i++){
for (int j=0; j<250; j++){
weight[i][j]=0;
heihgt[i][j]=0;
height[i][j]=0;
}
}
for (int i=0; i<A.train_set.size(); i++){
A.counts[A.train_set[i].label]++;
heihgt[A.train_set[i].label][int(A.train_set[i].height)]++;
// assert(int(A.train_set[i].height) >= 0 && int(A.train_set[i].height) < 250);
// assert(int(A.train_set[i].weight) >= 0 && int(A.train_set[i].weight) < 250);
if (A.train_set[i].height>249) A.train_set[i].height = 249;
if (A.train_set[i].weight>249) A.train_set[i].weight = 249;
height[A.train_set[i].label][int(A.train_set[i].height)]++;
weight[A.train_set[i].label][int(A.train_set[i].weight)]++;
}

Expand All @@ -99,7 +160,7 @@ Classifier barChart(Classifier A){
for (int i=0; i<2; i++)
for (int j=0; j<250; j++){
A.pWeight[i][j] = double(weight[i][j])/double(A.counts[i]);
A.pHeight[i][j] = double(heihgt[i][j])/double(A.counts[i]);
A.pHeight[i][j] = double(height[i][j])/double(A.counts[i]);
}
A.pWeight[0][0]=eps;
A.pWeight[1][0]=eps;
Expand Down Expand Up @@ -166,12 +227,15 @@ Classifier parzanRozenblatt (Classifier A){
if (A.counts[0] == 0 && A.counts[1]==0)
for (int i=0; i<A.train_set.size(); i++)
A.counts[A.train_set[i].label]++;
A.p[0] = double(A.counts[0])/double(A.train_set.size());
A.p[1] = double(A.counts[1])/double(A.train_set.size());
srand(time(NULL));
int c = (int)(A.train_set.size()/2);
// int c = (int)(A.train_set.size()/10);
int c = 30;
int count = 0;
double countTMP = 0;
double tmpH = A.H;
double const eps = 0.01;
double const eps = 0.1;
for (int i=0; i<(10.00-A.H)/eps; i++){
countTMP = 0;
for (int j = 0; j < c; j++){
Expand All @@ -189,12 +253,92 @@ Classifier parzanRozenblatt (Classifier A){
return A;
}

Classifier KNN(Classifier A){
A.mathModel = 4;
random_shuffle(A.train_set.begin(), A.train_set.end());
int c = int(A.train_set.size()/10);
vector<pair<double, int>> *mydata;
mydata = new vector<pair<double, int>> [c];
for (int i=0; i<c; i++){
for (int j=0; j<A.train_set.size(); j++){
if (i!=j) {
mydata[i].push_back(make_pair(sqrt(pow(A.train_set[i].height-A.train_set[j].height, 2)+pow(A.train_set[i].weight-A.train_set[j].weight, 2)), A.train_set[j].label));
}
}
sort(mydata[i].begin(), mydata[i].end());
}
int **ans;
*ans = new int [c];
for (int i=0; i<c; i++)
ans[i]= new int [15];
for (int i=0; i<c; i++){
for (int j=0; j<15; j++){
ans[i][j] = findAns(j, mydata[i]);
}
}
int mycounts[15];
int tmpK = 0;
int max = 0;
for (int i=1; i<15; i++){
mycounts[i]=0;
for (int j=0; j<c; j++){
if (ans[j][i] == A.train_set[j].label) mycounts[i]++;
}
if (mycounts[i]>max) {
max = mycounts[i];
tmpK = i;
}
}
A.k = tmpK;
return A;
}

Classifier parzanRozenblatt2 (Classifier A){
// 1 X-Xi
//P(x)=--------- * sum K(-----)
// n*pi*h^2 h
//
// sqrt((x-xi)^2+(y-yi)^2)
//K(x)=----------------------- if sqrt((x-xi)^2+(y-yi)^2) < h
// pi*h^2
//
const double pi = 3.14159265;
A.mathModel = 5;
if (A.counts[0] == 0 && A.counts[1]==0)
for (int i=0; i<A.train_set.size(); i++)
A.counts[A.train_set[i].label]++;
A.p[0] = double(A.counts[0])/double(A.train_set.size());
A.p[1] = double(A.counts[1])/double(A.train_set.size());
// int c = (int)(A.train_set.size()/10);
int c = 30;
int count = 0;
double countTMP = 0;
double tmpH = A.H;
double const eps = 0.1;
for (int i=0; i<(10.00-A.H)/eps; i++){
countTMP = 0;
for (int j = 0; j < c; j++){
//сократим на 1/pi*h^2
if (kernel1(tmpH, j, 0, A.train_set)/A.counts[0] >= kernel(tmpH, j, 1, A.train_set)/A.counts[1] && A.train_set[j].label == 0) countTMP++;
if (kernel1(tmpH, j, 1, A.train_set)/A.counts[1] > kernel(tmpH, j, 0, A.train_set)/A.counts[0] && A.train_set[j].label == 1) countTMP++;
}
if (countTMP > count){
count = countTMP;
A.H = tmpH;
}
tmpH+=eps;
}

return A;
}

void Classifier::train(ifstream &F, Classifier (*f)(Classifier A)){
readTrainFile(F);
train(this->train_set, f);
}
void Classifier::train(vector<d> input, Classifier (*f)(Classifier A)){
*this = f(*this);

}

vector<int> Classifier::classify(ifstream &F){
Expand All @@ -203,11 +347,23 @@ vector<int> Classifier::classify(ifstream &F){

vector<int> Classifier::classify(vector<pair<double, double>>input){
vector<int>rez;
if (mathModel == 4){
vector<pair<double, int>> mydata;
for (int i=0; i<input.size(); i++){
for (int j=0; j<train_set.size(); j++){
mydata.push_back(make_pair(sqrt(pow(train_set[j].height-input[i].first, 2)+pow(train_set[j].weight-input[i].second, 2)), train_set[j].label));
}
sort(mydata.begin(), mydata.end());
rez.push_back(findAns(13, mydata));
mydata.clear();
}
} else {
for (int i=0; i<input.size(); i++){
double a = probability(input[i].first, input[i].second, 0);
double b = probability(input[i].first, input[i].second, 1);
if (a > b) rez.push_back(0);
else rez.push_back(1);
}
}
return rez;
}
8 changes: 7 additions & 1 deletion classifier.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,8 +24,9 @@ class Classifier{
double p[2];
double pWeight[2][250], pHeight[2][250];
int counts[2];
double ExW[2], DxW[2], ExH[2], DxH[2]; //матожидание&дисперсия
double ExW[2], DxW[2], ExH[2], DxH[2];//матожидание&дисперсия
double H; //коэфф сглаживания
int k;//knn
vector<d>train_set;

void readTrainFile(ifstream &F);
Expand All @@ -35,7 +36,10 @@ class Classifier{
public:
Classifier();
~Classifier();
Classifier(const Classifier &A);
friend Classifier barChart(Classifier A);
friend Classifier KNN(Classifier A);
friend Classifier parzanRozenblatt2(Classifier A);
friend Classifier normalDistribution (Classifier A);
friend Classifier parzanRozenblatt (Classifier A);
void train(ifstream &F, Classifier (*f)(Classifier A));
Expand All @@ -46,5 +50,7 @@ class Classifier{
Classifier barChart(Classifier A);
Classifier normalDistribution (Classifier A);
Classifier parzanRozenblatt (Classifier A);
Classifier KNN(Classifier A);
Classifier parzanRozenblatt2(Classifier A);

#endif /* classifier_hpp */
46 changes: 13 additions & 33 deletions main.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,40 +14,20 @@ using namespace std;
int main() {
Classifier my;

vector<pair<double, double>> myTest;
myTest.push_back(make_pair(190, 90));
myTest.push_back(make_pair(185, 86));
myTest.push_back(make_pair(165, 46));
myTest.push_back(make_pair(164, 49));
myTest.push_back(make_pair(179, 76));
myTest.push_back(make_pair(155, 42));

vector<d>test1;

ifstream F;
F.open("/Users/anastasiapopova/Desktop/YANDEX/lab2_Bayes/lab2/train_test_set");

my.train(F, barChart);
vector<int>ans = my.classify(myTest);
for (int i=0; i<ans.size(); i++){
cout<<ans[i]<<" ";
}
cout<<endl;

my.train(F, normalDistribution);
ans = my.classify(myTest);
ifstream F, F1;
ofstream A;
F.open("/Users/anastasiapopova/Desktop/YANDEX/lab2_Bayes/lab2/train_set.txt");
F1.open("/Users/anastasiapopova/Desktop/YANDEX/lab2_Bayes/lab2/test_set.txt");

A.open("/Users/anastasiapopova/Desktop/YANDEX/lab2_Bayes/lab2/new");
my.train(F, parzanRozenblatt2);
cout<<"train"<<endl;
vector<int>ans = my.classify(F1);
cout<<"classify"<<endl;
for (int i=0; i<ans.size(); i++){
cout<<ans[i]<<" ";
A<<ans[i]<<endl;
}
cout<<endl;

my.train(F, parzanRozenblatt);
ans = my.classify(myTest);

for (int i=0; i<ans.size(); i++){
cout<<ans[i]<<" ";
}
cout<<endl;

A.close();

return 0;
}