Model evaluation

Hello everyone, I have a doubt, does the “numeric score” node evaluate the training or test data?

that depends on the input to the predictor. In your case, it is the test data as that is what is coming out of the bottom port of the Partitioning node. The Regression Predictor creates a prediction for each row in that data and then gives the result to the Numeric Scorer, which calculates R^2, etc. If you’d connect the top port of the partitioning node to the Regression Predictor, your Scorer would calculate the metrics on the training data, which of course does not make that much sense most of the time.
Kind regards,


Thank you very much Alexander

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.