In my research, I have split the test data set in order to handle my problem with svm. So I run the algorithm many times. Now I would like to visualize my results with PR and Roc curve. Is there an way to combine all the results to plot these curves?
PS. To plot the PR curve we use the confidence array and setting varius thresholds we get differents results (I have 2-class problem).
Re: Combine results and visualize PR and ROC curves
with PR curve you mean precision plotted versus recall? Do you really need that, if you already have the ROC curve? In any case, you can create the data for both curves with the technique of iterating different confidence values. If the ROC plot alone is sufficient, you can also simply use the Performance (Binominal Classification) operator to calculate the AUC measure and then view the ROC curve in the results view of the performance.