A Behavior-Item Based Hybrid Intention-Aware Frame for Sequence Recommendation

Sequence recommendation is one of the hotspots of recommendation algorithm research. Most of the existing sequence recommendation methods focus on how to use the items’ attributes to characterize the user’s preferences, ignoring that the user behavior also can reflect the preference for items. However, user behavior often has problems of mis-interaction and random interaction, which leads to fully utilizing it difficultly. Therefore, this paper proposes a new Behavior-Item based Hybrid Intent-aware Framework (BIHIF). In this framework, the user’s main intent is extracted based on user behaviors and interactive items, respectively, the two intent vectors are combined and extracted by the full connection layer to obtain the user’s real intent. We use real intent and item vector to calculate the score of the candidate items and make Top-K recommendations. Based on the framework, we implement models respectively by MLP and GRU, which show good results in the experiments based on three real-world datasets.