clear all
close all
load vibrate23;
Sample=a;
[row column]=size(Sample);
value0=Sample(:,column);
standdata=Sample;
rate=[];
x=[];
m=0;
noise=randn(20,8);
%load noise9;
for n=-20:0.1:20
m=m+1;    
standdata=[Sample;(Sample(:,1:column-1)+n*noise),value0];
%排序
[value,order]=sort(standdata(:,column));
for i=1:column-1
    datai=standdata(:,i);
    orderi=datai(order);
    standdata(:,i)=orderi;
end





catenu=max(value);
classnu=[];
trainnu=[];
beginloc=[1];
overnu=0;
overnut=0;
overloc=[];
overloct=[];
Samples1=[];
Labels=[];
testS1=[];
testL=[];
for i=1:catenu
  number=length(find(value==i));
  classnu=[classnu,number];
  trainnu=[trainnu,round(number*2/3)];
  overnut=overnut+number;
  beginloc=[beginloc,overnut+1];
  overnu=beginloc(i)+round(number*2/3)-1;
  overloc=[overloc,overnu];
  overloct=[overloct,overnut];
 
  Samples1=[Samples1;standdata(beginloc(i):overloc(i),1:column-1)];
  Labels=[Labels;value(beginloc(i):overloc(i))];
  testS1=[testS1;standdata(overloc(i)+1:overloct(i),1:column-1)];
  testL=[testL;value(overloc(i)+1:overloct(i))];
   
end
%catenu
%pause
%tic
[AlphaY1, SVs1, Bias1, Parameters1, nSV1, nLabel1] =LinearSVC(Samples1', Labels');
%toc
%nSV1
[ClassRate1, DecisionValue1, Ns1, ConfMatrix1, PreLabels1]= SVMTest(testS1',testL', AlphaY1, SVs1, Bias1,Parameters1, nSV1, nLabel1);
%ClassRate1
rate(m)=ClassRate1;
x(m)=n;
x=DecisionValue1;
mfparams = [log(.25) 0];
mftype = 'sigmf';
y=evalmf(x,mfparams,mftype);
plot(x,y,'.','r')
grid on

%ConfMatrix1
end
stairs(x,rate,'o')


figure(1)
[Labels, DecisionValue]= SVMClass(Samples1', AlphaY1, SVs1, Bias1, Parameters1, nSV1, nLabel1);
plot(Labels,DecisionValue,'.')
hold on
[Labels, DecisionValue]= SVMClass(testS1', AlphaY1, SVs1, Bias1, Parameters1, nSV1, nLabel1);
plot(Labels,DecisionValue,'o')


[Labels, DecisionValue]= SVMClass(SVs1, AlphaY1, SVs1, Bias1, Parameters1, nSV1, nLabel1);
x=DecisionValue;
mfparams = [log(.25) 0];
mftype = 'sigmf';
y=evalmf(x,mfparams,mftype);
plot(x,y,'.','r')
grid on
hold on