Performance (Binominal Classification) Error: Incompatible Label and Prediction

amitd

I have a simple example for demonstrating the effect of varying threshold on error, accuracy, sensitivity, etc. It works fine when I select classification error and the accuracy as metrics. However, the moment I add any of the metrics like precision, recall, AUC, I get the error in the Performance (Binominal Classification) operator: 
"Incompatible Label and Prediction" with the following details: The prediction attribute's values do not match the label attribute's. If the prediction attribute's values do not match the label attribute's, performance cannot be calculated.
While I understand what the error is trying to inform, there is no indication that the prediction attribute's values do not match those of the label's. I have checked this with breakpoints, etc. I have attached the data CSV file, working process RMP file and the erroneous process RMP file.

@sgenzer or anyone else can help?

rdesai

The problem is when you want to work with Recall and Precision, it expects the  value to be same for both the attributes(label(class) and prediction). In your case, you have set the threshold 0.0 and you have selected "fail on error" for optimzation grid operator.  so right now class and prediction column value(owner and true) do not match with your class column(nonowner and owner).Please see the image below. 
So, instead try setting different threshold value or select "ignore error" for optimzation grid operator. Then your class and prediction columns will look like this. 
and the result of the binomial classification including Recall and precision will look like this for each iteration 


Hope this helps!

rdesai

The problem is when you want to work with Recall and Precision, it expects the  value to be same for both the attributes(label(class) and prediction). In your case, you have set the threshold 0.0 and you have selected "fail on error" for optimzation grid operator.  so right now class and prediction column value(owner and true) do not match with your class column(nonowner and owner).Please see the image below. 
So, instead try setting different threshold value or select "ignore error" for optimzation grid operator. Then your class and prediction columns will look like this. 
and the result of the binomial classification including Recall and precision will look like this for each iteration 


Hope this helps!