I built a logistic regression on the entire sample (first model), as well as regressions on separate pieces (disjoint) of same sample (set of models). What is the best way to compare the effectiveness of the first model and the set of models (which are built on segments)? A common practice for evaluating the effectiveness of a set of models is some variant of the average AUCROC/GINI (does not matter) (as, for example, in the roc_auc_score for a multi-label case). Can it be permissible to calculate AUCROC for a union of predict_probas from the segments? And then compare that AUCROC vs AUCROC of first model?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.