"ROC analysis module for various classifiers"

ivalavan

Dear all,

I am trying to use the ROC analysis module within a RM process with embedded learning modules a knn, an SVM and neural network (one at at a time)
Keeping the default  configuration, a rather strange ROC i.e.. the same for all classifiers is outputed (a straight line with same TP =0.5 for all FP values).
When all three learning modules are embedded no learning rate is shown!

Has anyone similar experience or any experience on the ROC analysis module?

Which parameters are changed for all other classifiers than neural network? ( I assume that for neural neural network the output neuron threshold is changed and TP, FP values are calculated). Where could I find such information specifically for RM?

I would appreciate a prompt answer.

Thanks,

I.V.