Say we have a recommender system in production which recommends 1 our of N items according to some internal algorithm f given inputs Xi for each user i, let's assume f is a black box model. We have historical data recording user's reaction yi (click, dismiss, conversion, could be delayed), to the recommendation.

How should I evaluate the performance of this recommender system offline given the data available?


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.