LSSVM LAB Performance Metrics

Hello,

I am using LSSVMLAB to implement Least Square Support Vector Machine Algorithm.

I need to determine the accuracy, FPR, FNR, and other performance metrics results. How can I find them?

This is a code, getting some performance metrics, but it gets more than 1 value for each! It gives the result for each threshold.(For example, for 100 thresholds in ROC, it gets 100 TP results)

[area, se, thresholds, oneMinusSpec, sens, TN, TP, FN, FP] = roc(Y_latent,Y);

Answer 1

[maryam hatami]

I found the answer. All metrics are evaluated from TP, FP, TN, FN getting from the data:

len =length(Yt);
TP = 0;
FP = 0;
TN = 0;
FN = 0;
for i=1:len
   if Yt(i)==Yvalid(i)==1
     TP=TP+1;
   end
   if (Yvalid(i)==1) && (Yt(i)~=Yvalid(i))
     FP=FP+1;
   end
   if (Yt(i)==Yvalid(i)==0)
     TN=TN+1;
   end
   if (Yvalid(i)==0) && (Yt(i)~=Yvalid(i))
     FN=FN+1;
   end
end
TN
FP
FN
TP
fprintf(1,'Precision: %2.2f\n',100*sum(TP)/(TP+FP));
fprintf(1,'Recall: %2.2f\n',100*sum(TP)/(TP+FN));
fprintf(1,'F1_Score: %2.2f\n',2*(prc*recall)/(prc+recall));
fprintf(1,' FPR: %2.2f\n',100*(FP)/(TN+FP));
fprintf(1,' FNR: %2.2f\n',100*(FN)/(TP+FN));
fprintf(1,' TNR: %2.2f\n',100*(TN)/(TN+FP));
fprintf(1,'Accuracy: %2.2f\n',100*(TP+TN)/(TN+TP+FN+FP));

[Amr Embaby]

i'm trying to use robustLssvm function but i get that error :

SWITCH expression must be a scalar or a character vector.

Error in weightingscheme (line 9)

switch wfct Error in Roboustlssvm (line 74)

w = weightingscheme(casses,model.weights,delta);

do you know how to solve it?

my code:

type = 'function estimation';

[Yp,alpha,b,gam,sig2,model] = lssvm(fts_tle_si_x_wz_2_tra_nrm_mtx,tgs_tle_si_x_wz_2_tra.b,type);

model = tunelssvm(model, 'gridsearch','leaveoneoutlssvm', {'mse'});

model = robustlssvm(model);