Research And Application Of Dynamic Collaborative Filtering Algorithm Based On Trust Information

The development of 5G technology has provided convenience to people's lives,and the problem of information explosion has followed.The emergence of the recommendation system has become one of the most important technologies to solve the problem of information explosion,and has received widespread attention in the industry.Collaborative filtering recommendation algorithm is one of the most successful personalized recommendation technologies,but the traditional collaborative filtering algorithm generally has the problems of data sparsity and cold start,and usually cannot respond to the change of users' interest in time,which affects the recommendation quality.Trust relationship is an indispensable part of people's life.Research by the famous Nielsen survey agency in the United States shows that about 90% of users tend to trust recommendations from friends.Therefore,many scholars introduce trust relationship as auxiliary information into collaborative filtering algorithm,which can effectively alleviate the problem of algorithm data sparsity and cold start.In addition,as the most intuitive factor to reflect the change of user preferences,timing characteristics are also widely used in collaborative filtering algorithm recommendations.This article focuses on the problems in collaborative filtering algorithm(CF)recommendation algorithms.The main work of this paper is as follows: Considering the characteristics of user preference drifting over time,with the help of the Ebbinghaus forgetting curve to describe the time decay characteristics,proposed a probabilistic matrix factorization algorithm based on Ebbinghaus forgetting curve(Ebbinghaus-Probabilistic Matrix Factorization,E-PMF).Comparative experimental results show that the time series feature can improve the recommendation accuracy of the collaborative filtering algorithm to a certain extent.Aiming at the sparseness problem in the traditional collaborative filtering algorithm,this paper uses trust information as an auxiliary parameter to create a user information matrix,and uses the trust transfer mechanism in the trust network to establish a trust model to quantify the trust between trustees.At the same time,it introduces time series factors that affect personal preference information,integrates trust model scoring matrix and time series information,and proposes a trust-based dynamic collaborative filtering algorithm(Trust-Dynamic Collaborative Filtering,TDCF).Through experimental verification,the algorithm improves the accuracy of 12.01% compared with the traditional collaborative filtering algorithm,and can effectively alleviate data sparsity and cold start problems.The TDCF algorithm is applied to the movie recommendation system,which can complete user information registration,movie rating and movie recommendation.In this process,by mining the display or implicit information,constructs the user preference model.The proposed algorithm model is used to give the recommendation result,and processed according to the existing evaluation information of the movie,and finally the Top-N movie recommendation result is obtained.