A Recommendation Model Based on Heterogeneous Information Network Interaction
Zongli Jiang, Jingwen Li, Nan Ma · 2021 17th International Conference on Computational Intelligence and Security (CIS) · 2021
Most traditional recommendation systems are based on homogeneous networks. However, a large number of real-world graphs or networks are inherently heterogeneous, involving a diversity of node types and relationship types. Although the existing recommendation approaches based on heterogeneous information network have achieved performance improvement, while they still face the following problems. On one hand, existing models usually define multiple meta-paths in heterogeneous graphs to capture the composite relations and guide neighbor selection. However, these models usually discard intermediate nodes along the meta-path and cannot fully mine the interaction information of heterogeneous graphs. On the one hand, they do not take into account the time sequence between user and items. To address these two limitations, we propose a new model named Time-aware Interaction networks for Recommendation based on Heterogeneous information network (TIH4Rec) to improve performance. Specifically, TIH4Rec employs two major components, i.e., an interaction module that aggregates semantic information inter-meta-path and a time-aware module that incorporates time information. Experiments on two publicly heterogeneous graph datasets demonstrate the performance gains of TIH4Rec comparing with baselines.