Application of graph modeling and contrast learning in recommender system

Abstract

With the wide application of personalized recommender system in various fields, how to improve the accuracy and personalized level of recommender system has become a research hotspot. In this paper, a method of combining graph modeling and contrast learning is proposed to improve the performance of recommendation system by mining complex user project interaction and user preference. We first construct the user-project interaction graph, and extract the features of the graph structure by graph neural network (GNN) . In particular, graph convolution network (GCN) is used to update the node representation, and comparative learning is introduced to optimize the feature representation so as to improve the accuracy and personalization of recommendation. The experimental results show that the proposed method is superior to the traditional method in accuracy, recall and F 1 score. By analyzing the mechanism of combining graph modeling and contrast learning, this paper further expounds the theoretical basis and practical application of improving the performance of recommender system, and points out the limitations of existing methods and the future research direction.