matlab贝叶斯分类算法,如何使用贝叶斯（bayes）分类器实现分类

outPut(1:3,1:3)=0;      %判别矩阵的初始化

class1=[];

class2=[];

class3=[];

%生成二维正态分布的样本2 X N 维的矩阵  样本程序

%训练样本

load train1;

load train2;

load train3;

load test1;

load test2;

load test3;

%     train1=mvnrnd([1 1],[4 0;0 5],100)';   %2 X N

%     train2=mvnrnd([7 2],[7 0;0 4],100)';

%     train3=mvnrnd([2 7],[2 0;0 4],100)';

%     %测试样本

%     test1=mvnrnd([1 1],[4 0;0 5],100)';   %2 X N

%     test2=mvnrnd([7 2],[7 0;0 4],100)';

%     test3=mvnrnd([2 7],[2 0;0 4],100)';

%---------------------------------------------------%

%先验概率

P(1)=length(train1)/(length(train1)+length(train2)+length(train3));

P(2)=length(train2)/(length(train1)+length(train2)+length(train3));

P(3)=length(train3)/(length(train1)+length(train2)+length(train3));

%计算相关量  cov(X)：协方差矩阵      Ave:均值

%--------------------------------------------------------%

W1=-1/2*inv(cov(train1'));

W2=-1/2*inv(cov(train2'));

W3=-1/2*inv(cov(train3'));%

Ave1=(sum(train1')/length(train1))';%计算平均值(2维列向量，2*1)

Ave2=(sum(train2')/length(train2))';

Ave3=(sum(train3')/length(train3))';

w1=inv(cov(train1'))*Ave1;

w2=inv(cov(train2'))*Ave2;

w3=inv(cov(train3'))*Ave3;%2

w10=-1/2*Ave1'*inv(cov(train1'))*Ave1-1/2*log(det(cov(train1')))+log(P(1));

w20=-1/2*Ave2'*inv(cov(train2'))*Ave2-1/2*log(det(cov(train2')))+log(P(2));

w30=-1/2*Ave3'*inv(cov(train3'))*Ave3-1/2*log(det(cov(train3')))+log(P(3));

%-----------------------------------------------------------%

for i=1:3

for j=1:100

if i==1

g1=test1(:,j)'*W1*test1(:,j)+w1'*test1(:,j)+w10;

g2=test1(:,j)'*W2*test1(:,j)+w2'*test1(:,j)+w20;

g3=test1(:,j)'*W3*test1(:,j)+w3'*test1(:,j)+w30;

if g1>=g2&g1>=g3

outPut(1,1)=outPut(1,1)+1;

class1=[class1,test1(:,j)];

elseif g2>=g1&g2>=g3

outPut(1,2)=outPut(1,2)+1;%记录误判情况

class2=[class2,test1(:,j)];

else

outPut(1,3)=outPut(1,3)+1;%记录误判情况

class3=[class3,test1(:,j)];

end

elseif i==2

g1=test2(:,j)'*W1*test2(:,j)+w1'*test2(:,j)+w10;

g2=test2(:,j)'*W2*test2(:,j)+w2'*test2(:,j)+w20;

g3=test2(:,j)'*W3*test2(:,j)+w3'*test2(:,j)+w30;

if g2>=g1&g2>=g3

outPut(2,2)=outPut(2,2)+1;

class2=[class2,test2(:,j)];

elseif g1>=g2&g1>=g3

outPut(2,1)=outPut(2,1)+1;

class1=[class1,test2(:,j)];

else

outPut(2,3)=outPut(2,3)+1;

class3=[class3,test2(:,j)];

end

else

g1=test3(:,j)'*W1*test3(:,j)+w1'*test3(:,j)+w10;

g2=test3(:,j)'*W2*test3(:,j)+w2'*test3(:,j)+w20;

g3=test3(:,j)'*W3*test3(:,j)+w3'*test3(:,j)+w30;

if g3>=g1&g3>=g2

outPut(3,3)=outPut(3,3)+1;

class3=[class3,test3(:,j)];

elseif g2>=g1&g2>=g3

outPut(3,2)=outPut(3,2)+1;

class2=[class2,test3(:,j)];

else

outPut(3,1)=outPut(3,1)+1;

class1=[class1,test3(:,j)];

end

end

end

end

outPut

%---------------------------------------------------%

%画出各样本的分布情况

subplot(3,1,1)

plot(train1(1,:),train1(2,:),'go','LineWidth',2),hold on

plot(train2(1,:),train2(2,:),'b+','LineWidth',2),hold on

plot(train3(1,:),train3(2,:),'r.','LineWidth',2),hold on

title('训练样本分布情况')

legend('训练样本1','训练样本2','训练样本3')

subplot(3,1,2)

plot(test1(1,:),test1(2,:),'go','LineWidth',2),hold on

plot(test2(1,:),test2(2,:),'b+','LineWidth',2),hold on

plot(test3(1,:),test3(2,:),'r.','LineWidth',2),hold on

title('测试样本分布情况')

legend('测试样本1','测试样本2','测试样本3')

subplot(3,1,3)

plot(class1(1,:),class1(2,:),'go','LineWidth',2),hold on

plot(class2(1,:),class2(2,:),'b+','LineWidth',2),hold on

plot(class3(1,:),class3(2,:),'r.','LineWidth',2),hold on

title('测试样本分类后分布情况')

legend('测试样本1','测试样本2','测试样本3')