what is the metric used to gauge the significance of ranking using SPSS's MLP-NN ? (is there something like Matthew's coeficient in SPSS?)

Question

I have used SPSS's Multi-layer perceptron model to rank some variables according to their importance in contribution to a specified target.

My question is... what is the metric used to gauge the performance of the model?

In non-SPSS NN models, one would use something like the Matthew's coefficient to gauge the performance, is there a metric for the MLP-NN in SPSS?

Answer 1

The method used for computing the predictor importance in a predictive model is the same across many of the algorithms in IBM SPSS Modeler.

You can find the details in the chapter "Predictor Importance Algorithms" of the "IBM SPSS Modeler Algorithms Guide".

You can find a copy here.