Computing and Informatics

The traditional collaborative filtering recommendation model usually cannot capture the correlation between user items accurately when faced with the problem of data sparsity, which leads to a poor recommendation effect. To solve these problems, this paper proposed a recommendation model integrating semantic features and knowledge features. By introducing a knowledge graph as auxiliary information, the problem of data sparseness was alleviated, and the granularity of user interest description was refined based on the rich knowledge information. Different from other recommendation models based on knowledge graphs, the model in this paper mined implicit feedback information between user projects based on semantic features and knowledge features. The Word2vec model was used to capture the semantic features of project entities, and knowledge features were learned according to the idea of preference diffusion to obtain the user preference diffusion set integrating features of different levels and forming the user preference feature sequence from the inner layer to the outer layer. The sequence was adopted as the input of the gated cycle unit, and the attention mechanism was integrated to capture the deep feature information between the user and the item, so as to enrich the user representation to improve the recommendation effect. The comparison experiment of the CTR prediction scenario and the Top-K recommendaton scenario, and the sparseness verification experiment were all conducted on the MovieLens dataset. The results showed that, compared with BPR-MF, DKN and RippleNet baseline models, this model had better performance in both scenarios and could effectively alleviate the impact of data sparseness.