Probability output of K-Nearest-Keighbours ?


In the usage of the KKNeighbour algorithm for classification, I am interested to have as output the column of probabilities of correct classification. (ideally I would like to be able to define also the cutoff).

My reason is simple: I would like to measure the performance of the learned classifier by using  a ROC Curve or a lift curve.

It seems that the KNN-Knime does not show as output the probability column. Can you help me in this matter?



Due to the nature of the NN classifier, i see no trivial method for determining the probability, unlike the NB where the classification is directly based on probability or DT where one could distinguish certain bins and derive a ratio of correctly classified items.