Heterogeneous Knowledge Graph Personalized Recommendation Algorithm

Abstract

In recent years, the use of knowledge graph and deep learning for recommendation has been widely studied and applied, but most recommendation models are incomplete in the representation of heterogeneous information items. Aiming at the extraction of high-order representations of user heterogeneous features and project features, a personalized recommendation model based on heterogeneous knowledge graph (HKGR) is proposed. The propagation of heterogeneous neural networks is used to model user preferences and project characteristics to improve recommendation performance. The accuracy of model recommendation is improved, and the ability of the model to cope with data sparse scenarios is enhanced. The experimental results show that the key components in this method are compared with the co-field baseline method, which is conducive to improving the recommended performance.