A Collaborative Filtering Recommendation Method with Integrated User Profiles

In the article recommendation, text information as the main body of the recommendation is rich in semantic content. Especially for content-based recommendation methods, whether an accurate and concise feature representation can be extracted from existing text information is the key to the effective recommendation. Since the long-term use of content-based recommendation methods to generate personalized result sets can make the recommendation variety too homogeneous, the collaborative filtering recommendation method compensates for the above problem by finding other preferred articles of similar users for the recommendation. In this paper, we propose a collaborative filtering recommendation method that incorporates user profiles. This method designs a user portrait labeling system for the article recommendation scenario. Moreover, it uses relevant text processing techniques to extract multidimensional user features, which can alleviate the cold start and matrix sparsity problems when performing collaborative filtering recommendations. Finally, we tested our scheme with the MIND Data Set and analyzed the advantages of our scheme.