RippleNet network combines path and entity information, and is successfully applied to recommendation systems. Nevertheless, the model ignores the impact of the relationships between the entities. As a result, this paper introduces an enhanced RippleNet model that incorporates the impact of entity relationships into the preference propagation model by mapping the entity embedding into the relation vector space and integrates it into RippleNet as entity weights. And according to user historical data, preferences are propagated on the knowledge graph. The experimental findings manifest that the new approach enhances the AUC and ACC values of RippleNet, surpassing other baseline methods. These results show that our model effectively addresses the issue of the inter-entity relationship not being considered in the RippleNet network, resulting in recommendation outcomes that better align with user expectations.
A Knowledge Graph Preference Propagation Recommendation Algorithm Based on Introducing Relational Information
11.10.2023
2664694 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
British Library Online Contents | 2018
|Travel recommendation method based on knowledge graph and traffic speed prediction
Europäisches Patentamt | 2024
|