A Graph Collaborative Filtering Recommendation Method Based on Dual Message Propagation Mechanism

Abstract

The invention discloses a graph collaborative filtering recommendation method based on a dual-message propagation mechanism. The method designs a message propagation mechanism specifically for recommendation scenarios—the dual-message propagation mechanism. This mechanism expands the traditional message propagation mechanism, its advantage is that it can fully mine and utilize two important relationships between users and products at the same time; based on the dual message propagation mechanism, a new graph collaborative filtering recommendation method is designed, and two graph neural networks are used to process The vector representation of each user and product, in these two graph neural networks, one can complete the modeling of the preference relationship, and the other can complete the modeling of the similarity relationship, this method overcomes the existing graph collaborative filtering The shortcomings of the method improve the recommendation effect.

Description

technical field [0001] The invention belongs to a graph collaborative filtering recommendation method, in particular to a graph collaborative filtering recommendation method based on a dual message propagation mechanism. Background technique [0002] Existing graph collaborative filtering recommendation methods mainly use message propagation mechanism and graph neural network for recommendation. However, there is a serious problem in the existing graph collaborative filtering recommendation method: that is, graph neural network was first successfully applied to graph (graph) classification , graph (graph) representation and other fields closely related to graph (graph), these fields are quite different from the recommendation field. The existing graph collaborative filtering recommendation method directly imitates the use of graph neural network in the graph (graph) field, that is, directly uses the traditional message propagation mechanism to obtain and optimize the vector ...

AI Technical Summary

Problems solved by technology

[0002] Existing graph collaborative filtering recommendation methods mainly use message propagation mechanism and graph neural network for recommendation. However, there is a serious problem in the existing graph collaborative filtering recommendation method: that is, graph neural network was first successfully applied to graph (graph) classification , graph (graph) representation and other fields closely related to graph (graph), these fields are quite different from the recommended field

Due to failure to meet this premise, there is a bottleneck in the recommendation effect of the existing graph collaborative filtering recommendation method

Images