A Trust-Based Collaborative Filtering Algorithm Using A User Preference Clustering

With the rapid development of mobile Internet and big data technology,we are surrounded by vast amounts of information every day.How to effectively find the information we need in the vast amount of information is the problem we face today.Researchers have come up with a solution for a personalized recommendation system which can recommend useful information to us like our friends.Among them,the collaborative filtering recommendation algorithm has been widely studied and applied.However,the low scalability and data sparsity of the algorithm have become more and more obvious as the data size of users and projects on the Internet continues to expand.In order to solve the problem of data sparsity and scalability in collaborative filtering algorithms,from the perspective of reducing the search scope of neighboring users and the accuracy of similarity measures in the recommendation system,A trust-based collaborative filtering algorithm using a user preference clustering is proposed.First,users are clustered based on user preference through the K-means clustering algorithm.According to the project information,construct the project-type matrix,and calculate the user preference matrix according to the characteristics of user ratings.This process reduces the sparseness of the data and gets users' preference.Then we cluster users with K-means method to reduce the scope of searching for the nearest neighbor to a cluster of users with similar preferences and improve the computational efficiency and scalability of the algorithm.Then,we build a trust relationship model.Based on the original trust relationship model,the score evaluation factor and the user preference variable are integrated,and the direct and indirect trust degrees between users are calculated to obtain the user's overall trust level.Based on the similarity measurement method of trust relationship,the comprehensive similarity between users and the prediction rating are calculated.The Top-N project with the highest prediction score is recommended to the target user.Calculating the similarity with the trust relationship can improve the quality of the nearest neighbor's choice,so as to get a better recommendation effect.Simulation experiments were performed based on the Movie Lens classical data set to verify the validity of the model.After obtaining the optimal parameter values in the algorithm,four kinds of collaborative filtering algorithms were compared.Experimental results show that the improved algorithm can improve the efficiency of nearest neighbor query and the prediction accuracy of the score,relieve the scalability of the traditional collaborative filtering algorithm,and effectively improve the inaccuracy of the nearest neighbor selection for the traditional collaborative filtering algorithm in the case of sparse data.