Knowledge Graph-based Session Recommendation with Session-Adaptive Propagation

Session-based recommender systems (SBRSs) predict users' next interacteditems based on their historical activities. While most SBRSs capture purchasingintentions locally within each session, capturing items' global informationacross different sessions is crucial in characterizing their generalproperties. Previous works capture this cross-session information byconstructing graphs and incorporating neighbor information. However, thisincorporation cannot vary adaptively according to the unique intention of eachsession, and the constructed graphs consist of only one type of user-iteminteraction. To address these limitations, we propose knowledge graph-basedsession recommendation with session-adaptive propagation. Specifically, webuild a knowledge graph by connecting items with multi-typed edges tocharacterize various user-item interactions. Then, we adaptively aggregateitems' neighbor information considering user intention within the learnedsession. Experimental results demonstrate that equipping our constructedknowledge graph and session-adaptive propagation enhances sessionrecommendation backbones by 10study showing our proposed framework achieves 2existing well-deployed model at The Home Depot e-platform.