I have carried out experiment with 4 different features of nucleotides with three kernels in SVM. One of the feature with a kernel is having the best accuracy. I am in a dilemma of how to draw a ROC curve. I know ROC curve is FPR in x-axis and TPR in y-axis. Is it sufficient to calculate TPR and FPR for the best feature obtained and draw curve for the same. Need some suggestion on this..
You should draw TPR vs FPR or better yet sensitivity over specificity while varying a parameter or more in you classification. The ROC curve demonstrates several things one of which is the area under the curve which could be used as a measure of accuracy.
Thanks Mr Ahmed. That was indeed helpful. Need one more help. How the AUC was calculated is not given there. I have a formula but dont know if it is correct. Do u have some idea on that.
u can use the following link for the roc curve
http://www.alanfielding.co.uk/multivar/accuracy.htm
we can use phd toolbox matlab
http://www.mathworks.com/matlabcentral/fileexchange/35106-the-phd-face-recognition-toolbox
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix