Research On Collaborative Recommendation Algorithm Based On User Clustering

With the rapid development of information technology and Internet, especially the emergence of e-commerce sites and social networking sites in recent years, information overload is becoming a serious problem for users and suppliers. To resolve this problem, recommendation systems are proposed, which analyze the relationship between the user’s interest and the items, according to the user’s information, and provide personalized recommendation service for users. Collaborative filtering is a widely used technology of personalized recommendation systems. The traditional collaborative filtering algorithms are divided into user-based and item-based.According to the user’s history score, collaborative filtering algorithm based on user calculates the similarity between users, finds the nearest neighbor for the user, and produce the final recommendation. Due to the problem of data sparsity and cold start, accuracy of recommendation was seriously affected. On the other hand, with the increase of system users and project, the recommended time for user consumption will continue to increase, which will affect the scalability of the system and the efficiency of recommendation.Based on the problems of additional collaborative filtering, we proposed a collaborative filtering algorithm based on clustering of users and trust. The algorithm clusters users in the system background preprocessing stage, to reduce the system to search for nearest neighbor users, improves efficiency and system scalability. The algorithm introduces trust relationships between users in the recommendation process, combines the user’s rating similarity and user trust degree in order to improve the accuracy of recommendation and solves the cold start problem.Comparative experiments are carried out in the selected data sets.The experimental results show that the proposed UTCF can enhance the accuracy of system effectively in the case of user rating matrix sparsity.