Session-based Recommendation with Global Information Integration and Multi-level Intent Modeling

• 1、Beijing Jiaotong University, School of Computer Science and Technology, Beijing 100044

Abstract：Session-based Recommendationaims to predict the next item of interest for an anonymous user based on user\'s behavior in the current session.Although current research has achieved excellent results, there are still some problems. Most existing studies only use informaiton within a single session, ignoring the information in other sessions, while some multi-session models may introduce noise that affects the performance of the model when using cross-session information. Besides, due to the influence of early work, existing models only consider the last item in the session as the user\'s current intent, and do not take into account the user\'s multi-level intents, which may lead to inaccurate modeling of user intents. To solve the above problems, this paper proposes a Session-based Recommendation method that combines global information and multi-level intents. The model introduces global information into the current session by sampling neighbor nodes from the global graph and using the attention mechanism to reduce the interference of noise. At the same time, the model generates multi-level intent representation of the user through the multi-level intent module, which improves the model\'s ability of modeling user intents. Experimental results on three datasets demonstrate the effectiveness of the proposed method.

Keywords： Session-based Recommendation Graph Neural Networks Computer Application Technology