confusionmatrix.m

来自「各种SVM分类算法」· M 代码 · 共 29 行

function   [C,metric]=ConfusionMatrix(ypred,ytrue,classcode)
%  [C,metric]=ConfusionMatrix(ypred,ytrue,classcode)
%
%  ypred and ytrue -1, 1
%  if classcode = 1 -1
%
%  TP FN
%  FP TN
C=zeros(2);
N=length(classcode);
for i=1:N
    for j=1:N
    C(i,j)=  length(  find(ypred==classcode(j) & ytrue == classcode(i))); 
end;
end;
nbpos=sum(ytrue==1);
nbneg=sum(ytrue==-1);
c=nbneg/nbpos;
metric.detection=C(1,1)/sum(C(1,:));
metric.accuracy=sum(diag(C))/sum(sum(C));
if sum(C(:,1))~=0
metric.precision=C(1,1)/(C(1,1)+C(2,1));
else
    metric.precision=NaN;
end;
metric.fmeasure=2*C(1,1)/(C(1,1)+C(2,1)+nbpos);
metric.wracc=(4*c)/(1+c)^2*(C(1,1)-C(2,1));