Bayes判别matlab源程序

在分类判别中，bayes判别的确具有明显的统计优势，在我所做的应用中，与模糊，灰色，物元可拓相比，判别准确率一般都会高些，而BP神经网络由于调试麻烦，在调试过程中需要人工参与，而且存在明显的问题，局部极小点和精度与速度的矛盾，以及训练精度和仿真精度间的矛盾，等，尽管是非线性问题的一种重要方法，但是在我们项目中使用存在一定的局限，基于此，研究bayes判别，并写出bayes判别的matlab程序，与spss非逐步判别计算结果一致。现对外共享下：

%贝叶斯判别

clear;clc;

yangben=[

1 5.38 2.44 2.77 8.53 1.58 0.48

1 1.11 1.62 3.05 4.29 0.29 0.08

1 3.24 2.04 4.55 8.16 1.25 0.42

2 2.44 2.09 36.05 5.44 28.72 6.25

2 2.20 2.32 35.94 5.03 28.78 6.36

2 2.12 1.90 36.80 5.70 28.93 6.18

2 2.96 3.27 34.55 5.48 28.67 6.61

2 2.25 1.80 36.43 5.33 28.95 5.83

2 2.19 1.87 36.69 5.55 28.98 6.23

2 1.97 1.22 37.61 5.06 31.46 4.25

2 2.64 1.41 35.73 5.03 32.21 2.54

3 2.17 1.22 32.1 6.41 28.93 0.15

3 1.81 1.05 31.03 3.89 28.78 1.22

3 1.49 0.83 31.06 3.85 29.01 0.52

3 2.15 1 30.61 5.47 28.07 0.22

3 0.36 0.17 51.05 23.45 28.06 0.07

3 0.12 0.10 62.31 40.91 17.37 0.66

3 0.22 0.11 53.17 21.88 28.64 0.09

3 2.33 1.64 33.95 7.66 29.54 0.72

3 0.93 0.41 11.71 6.49 1.6 4.96

];

g=3; %g-group分类数，后来写了个自动检测分类数的，不过没在matlab下些，呵呵

[m,n]=size(yangben);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

for i=1:g

groupNum(i)=0;

group(i)=0;

for j=1:m

if yangben(j,1)==i

group(i)=group(i)+1;

end

end

if i==1

groupNum(i)=group(i);

else

groupNum(i)=groupNum(i-1)+group(i);

end

end

group;

groupNum; %计算分类个数数组

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%计算总平均值

% for j=1:n-1

% TotalMean(j)=0;

% for i=1:m

% TotalMean(j)=TotalMean(j)+yangben(i,j+1);

% end

% TotalMean(j)=TotalMean(j)/m;

% end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

GroupMean=[];

for i=1:g

if i==1

low=1;

up=groupNum(i);

else

low=groupNum(i-1)+1;

up=groupNum(i);

end

matrix=yangben(low:up,:);

MatrixMean=mean(matrix); %各分类组平均值

GroupMean=[GroupMean;MatrixMean];

for u=low:up

for v=2:n

C(u,v-1)=yangben(u,v)-MatrixMean(v);

end

end

end

C;

GroupMean;

V=C'*C/(m-g);

V_inv=inv(V); %对矩阵V求逆

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

GroupMean=GroupMean(:,2:n);

Q1=GroupMean*V_inv;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

for i=1:g

lnqi(i)=log(group(i)/m);

mat=GroupMean(i,:);

Q2(i)=lnqi(i)-0.5*mat*V_inv*mat';

end

lnqi;

Q2;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

b=[

5.38 2.44 2.77 8.53 1.58 0.48

1.11 1.62 3.05 4.29 0.29 0.08

3.24 2.04 4.55 8.16 1.25 0.42

3.12 1.78 3.47 7.41 0.74 0.22

3.22 1.91 3.87 7.69 0.97 0.34

1.91 1.68 34.6 6.03 27.47 4.69

2.77 1.84 34.74 5.58 29.36 4.41

2.5 1.55 37.47 5.45 30.71 5.36

2 1.09 35.2 4.59 30.27 3.43

2.62 2.25 35.56 5.68 28.96 5.79

3.23 2.93 32.86 5.43 27.6 5.37

2.44 2.09 36.05 5.44 28.72 6.25

2.20 2.32 35.94 5.03 28.78 6.36

2.12 1.90 36.80 5.70 28.93 6.18

2.96 3.27 34.55 5.48 28.67 6.61

2.25 1.80 36.43 5.33 28.95 5.83

2.19 1.87 36.69 5.55 28.98 6.23

1.97 1.22 37.61 5.06 31.46 4.25

2.64 1.41 35.73 5.03 32.21 2.54

0.14 0.19 69.64 41.52 13.73 0.05

0.14 0.16 18.68 16.45 0.58 0.31

0.21 0.4 78.97 42.36 27.18 0

0.3 0.2 77.93 38.32 36.05 0.64

0.32 0.29 56.87 21.26 36.2 0.02

0.1 0.16 50.53 16.45 30.38 2.29

0.06 0.22 50.48 13.46 31.05 1.78

0.19 0.09 35.69 6.5 28.34 0.77

0.26 0.24 35.41 6.88 28.37 0.66

0.13 0.15 40.05 10.43 28.36 1.54

1.26 0.39 35.16 7.97 24.85 3.99

0.88 0.53 38.46 7.76 26.24 5.87

1.27 0.57 37.06 7.84 27.26 3.8

0.22 0.22 43.18 11.88 28.93 2.81

0.15 0.13 53.09 15.67 35.11 0.08

1.28 0.61 34.37 0.39 28.94 1.92

0.33 0.6 37.65 6.38 30.05 0.97

0.49 0.31 53.97 15.4 32.87 3.69

2.17 1.22 32.1 6.41 28.93 0.15

1.81 1.05 31.03 3.89 28.78 1.22

1.49 0.83 31.06 3.85 29.01 0.52

2.15 1 30.61 5.47 28.07 0.22

0.36 0.17 51.05 23.45 28.06 0.07

0.12 0.10 62.31 40.91 17.37 0.66

0.22 0.11 53.17 21.88 28.64 0.09

2.33 1.64 33.95 7.66 29.54 0.72

0.93 0.41 11.71 6.49 1.6 4.96

]; % 判别的样本

[u,v]=size(b);

result=[];

for i=1:u

x=b(i,:);

yy=Q1*x'+Q2';

result=[result yy];

end

res=result' %计算的待判数据对各标准数据的线性计算值

[rows,cols]=size(result);

for i=1:cols

iljj=0;

mlljj=result(:,i);

for j=1:rows

iljj=iljj+exp(result(j,i)-max(mlljj));

end

for j=1:rows

houyangailv(j,i)=exp(result(j,i)-max(mlljj))/iljj;

end

end

H=houyangailv' %后验概率