Collect "Scorer" (confusion matrix) of a neural network predictor for each fold of a Cross validation

[Soufiane8413]

Hello, I would like to collect the confusion matrix and the AUC for each fold of the cross validation. The statistics, scorer, and ROC curve nodes seems to show the result for the last one only.
Do you have any idea ?
Thank you in advance !

[Soufiane8413]

I’ve found the solution for the confusion matrix (with the prediction table that show all the results) but i dont have acces to the probabilities (output of the Keras Network Executor) required for the ROC curve. Could someone help me ?

[ScottF]

Hi @Soufiane8413 -

It’s a little clunky maybe, but what about using a CSV Writer node (with append option enabled) to save the probabilities to a file in each iteration of the cross validation loop? You could read that file in - and manipulate it a bit - after the CV loop completes to provide input for the ROC curve.

[Soufiane8413]

Thank you very much for your reply @ScottF !