A Next Basket Recommendation Reality Check

The goal of a next basket recommendation (NBR) system is to recommend items for the user, based on sequence their prior baskets. We examine whether performance gains NBR methods reported in literature hold up under fair and comprehensive comparison. To clarify mixed picture that emerges from our comparison, we provide novel angle evaluation methods, centered distinction between repetition exploration: typically composed previously consumed (i.e., repeat items) new explore items). propose set metrics measure repetition/exploration ratio models. Using these metrics, second analysis state-of-the-art results help extent actual progress achieved by existing as well underlying reasons any improvements observe. Overall, work sheds light problem NBR, provides protocol, yields useful insights design models this task.