Mathimatics-Numerical algorithms

Development Platform:

Matlab

cushing_svm.m：Code Content

							%============================
%使用SVM实现三类模式识别
%============================
clear;
clc;
load train.data;
train_data=train(:,1:2);% 全部训练样本集合
expected_output=[-1;-1;-1;-1;-1;-1;1;1;1;1;1;1;1;1;1;1;-1;-1;-1;-1;-1]; % 先分成ac和b两类时的全部期望输出
train_data_partly=[];
temp1=train_data(1:6,:);
temp2=train_data(17:21,:);
train_data_partly=[temp1;temp2];% 部分训练样本集合，a类和c类样本集合
expected_output_partly=[-1;-1;-1;-1;-1;-1;1;1;1;1;1];% 部分期望输出，a类和c类样本的期望输出集合
ker='rbf';
C=100.0;
X=train_data;
Y=expected_output;
[nsv alpha bias]=svc(X,Y,ker,C);
alpha1=alpha;
X=train_data_partly;
Y=expected_output_partly;
[nsv alpha bias]=svc(X,Y,ker,C);
alpha2=alpha;
load test.data
test_data=test;% 测试样本集
       
clc;
% 测试识别输出
fprintf('测试样本的识别结果为： n');
for i=1:6
    X=train_data;
    Y=expected_output;
    Z=test_data(i,:);
    predictedY = svcoutput(X,Y,Z,ker,alpha1,bias,0);
    if  (predictedY==1)
        fprintf(' b n');
    else
        X=train_data_partly;
        Y=expected_output_partly;
        predictedY = svcoutput(X,Y,Z,ker,alpha2,bias,0);
        if (predictedY==1)
           fprintf( ' c n');
        else
           fprintf(' a n');
        end
   end
end
    

						