Context-Aware Intent-Driven Session-based Model for Session-based Recommendation

This Session-based recommendation systems (SBR) serve a vital function in the digital information era. It primarily predicts the user's current session behavior and interactions with certain items. Currently, existing SBR models typically model only the user's preferences in the current session. They focus on the last item of the session. These models cannot provide a unified statement for the session problem. They also cannot promptly capture the multiple intents that may arise from the user during the session. This leads to a significant decline in recommendation performance, and these issues are more pronounced in longer sessions. To address the above issues. This paper adopts a novel session-based recommendation model (CISBR). It collects information about repeated items in the session through frequency embedding vectors. It derives information around each item across multiple sessions. It dynamically selects the most important representations of multiple user intents. The paper demonstrates through experimental results that CISBR outperforms existing Session-based recommendation systems models by a significant margin across several datasets. It also performs excellently on datasets with longer sessions.