A Context-Aware Collaborative Filtering Recommender System Based on GCNs

In order to solve the problem of data sparsity inherent in collaborative filtering systems, this paper proposes a context-aware collaborative filtering method based on user similarity measures. This method can effectively integrate context information, social information and other information into the recommendation process. First, a graph convolutional network (GCN) is used to learn the relationship between the context information of each user node and the user evaluation, and consider the potential social network structure in the recommendation system. Secondly, the direct similarity between neighborhoods is measured by representing the multi-source data as user scores of popular items, and then the indirect similarity between non-neighborhood users is calculated based on the similarity propagation. Finally, two similarity values are used to normalize the content in the user item scoring matrix to prevent overfitting. Experimental results on actual data sets show that this method can effectively alleviate the problem of data sparsity in the collaborative filtering system, and can integrate multi-source data, effectively improving the accuracy of the recommendation system.