Customer requests a feature to have an option for the training process to provide feedback on which data files caused false alarms. 

Currently at the end of training we display the confusion matrix that shows the predicted result vs the expected results for the validate file set. 

The request is to also identify the files whose predicted values were not what was expected.