Closed abellogin closed 10 years ago
Added a flag to not give too much advantage to recommenders with low coverage (if the flag is 1, then precision@k is normalised by k always).
Thanks @abellogin, fixed in release-0.2.2, with default behaviour now being to normalize by k and assume that all missing recommendations were incorrect.
Added a flag to not give too much advantage to recommenders with low coverage (if the flag is 1, then precision@k is normalised by k always).