Improved Recommendation Model Based on Knowledge Graph and Convolutional Networks
LIU Huijing
LI Ting
Abstract：The emergence of knowledge graphs has brought profound changes to recommender systems,making recommenda-tions interpretable while providing rich semantic information,and human personalities change over time,a factor that extant tech-niques neglect to bring to recommender systems.So this paper proposes a model that fuses time decay and knowledge graph convolu-tional networks(KGCN-TD),which introduces the time decay factor into the knowledge graph convolutional network model(KGCN),which can help the model capture the temporal nature of user behaviors,and takes the time information as an attribute of the edges in the knowledge graph,and converts the time into a weight by defining a decay function,which focuses on mining knowl-edge graph related attributes to capture item relevance,and extends the sensory field to capture the user's potential distant interests by sampling entity neighbor information and computing entity representations in combination with bias.This strategy respects the temporal order and preserves the structural information of the knowledge graph to further improve the prediction performance of KGCN.The model is effectively validated on three datasets,which are movies,books,and music.
Keywords：recommendation systemknowledge graphgraph convolutional networkstime decay factor
Publication Date：2025-11-20
Online Publishing Date：2026-01-28(First online date of this platform, not the publication date of the document)
Pages：5( 41-44,81 )