Collaborative Filtering Recommendation Based On Compound Information

Today Internet technology develops rapidly and information data comes out explosively.The way to make people get the effective information quickly becomes the focus of attention.The emergence of recommender system is an effective solution to this problem.Normally,the most widely used recommendation algorithm is collaborative filtering recommendation algorithm.The collaborative filtering recommendation algorithm is the method,which uses users' historical behaviors and users' feedback information to predict user preferences.However,problems of data sparseness and cold start seriously impact the effect of traditional recommendation.Therefore,this paper proposes a collaborative filtering recommendation method based on compound information.The use of compound information is mainly divided into two aspects: on the one hand,use the idea of cross domain recommendation.The cross domain recommendation method is to combine the data from multiple domains together to recommend in the target domain.Transfer information from the auxiliary domain to the target domain to improve the accuracy of recommendation in the target domain.On the other hand,different from only mining features from the rating information,this paper learns three kinds of explicit and implicit information from the rating information and tag information between users and items,trust information between users and users jointly.What's more,the algorithm is based on the SVD++ matrix decomposition algorithm.Through learning jointly of these three kinds of explicit and implicit information from auxiliary domain and target domain,we extend the user latent factors with tag latent factors and trust latent factors,and then combine with the item latent factors to make a prediction for the ratings of the target domain.Thus,the model alleviates the problems of sparse data and cold start,and achieves the purpose of improving the accuracy of the recommendation.The results of experiments show that the proposed collaborative filtering algorithm based on compound information is better than or close to the classical and recently proposed algorithms.Moreover,this algorithm can effectively improve the accuracy of the recommendation in the case of sparse data and cold start.