fisher.m

模式识别：线性分类器

clc;
clear all;

%load train set
load maleTrain
load femaleTrain

%define prior-probability
Pmale = 0.5;
%Pmale = 0.75;
%Pmale = 0.9;
Pfemale = 1-Pmale;

[Rmale,Cmale] = size(male);
[Rfemale,Cfemale] = size(female);

%the mean vector
meanMale = sum(male)/Cmale;
meanFemale = sum(female)/Cfemale;

%define with-class scatter matrices of male and female
SmaleHeight = male(:,1)-meanMale(1);
SmaleWeight = male(:,2)-meanMale(2);
Smale = [SmaleHeight , SmaleWeight];
Smale = Smale' * Smale;

SfemaleHeight = female(:,1)-meanFemale(1);
SfemaleWeight = female(:,2)-meanFemale(2);
Sfemale = [SfemaleHeight , SfemaleWeight];
Sfemale = Sfemale' * Sfemale;

%define with-class scatter matrices of human being
Sw = Pmale*Smale + Pfemale*Sfemale;

%define between-class scatter matrices
Sb = Pmale*Pfemale*(meanMale-meanFemale)*(meanMale-meanFemale)';

%compute the optimal weight vector solution
Wopt = inv(Sw)*(meanMale-meanFemale)';

%define the threshold
%y0 = 0 ;

y0 = Pmale*sum(male * Wopt)/50 +Pfemale*sum(female * Wopt)/50;


%load test set
%训练集描述，第三列表示性别，0为女，1为男

%load test1
load test1
out1 = test1(:,[1,2]) * Wopt;
[Rout1,Cout1] = size(out1);

for i = 1:Rout1
    if(out1(i)>=y0)
        out1(i) = 1;
    else
        out1(i) = 0;
    end
    
    if(out1(i) == test1(i,3))
        display('true');
        trueTable1(i) = 1;
    else
        display('false');
        trueTable1(i) = 0;
    end
end

%load test2
load test2
out2 = test2(:,[1,2]) * Wopt;
[Rout2,Cout2] = size(out2);

for i = 1:Rout2
    if(out2(i)>=y0)
        out2(i) = 1;
    else
        out2(i) = 0;
    end
    
    if(out2(i) == test2(i,3))
        display('true');
        trueTable2(i) = 1;
    else
        display('false');
        trueTable2(i) = 0;
    end
end

%plot the true table
figure;
stem(trueTable1,'r');
axis([0,Rout1,0,1.5]);
title('test1在fisher准则下判断的结果');
legend('输出结果的正确率，1表示判断正确，0表示判断错误');

figure;
stem(trueTable2,'r');
axis([0,Rout2,0,1.5]);
title('test2在fisher准则下判断的结果');
legend('输出结果的正确率，1表示判断正确，0表示判断错误');