Research On Personalized Recommendation Algorithm Based On Time Forgetting And Clustering

With the gradual popularization of 5G and the accelerated development of the Internet,network information has exponentially increased.Confronted with a substantial number of data,how users can find the information they need,and how information providers make the information they provide to be paid attention to and collected by target users have become urgent issues that need to be resolved.Based on this,the recommendation system emerged as a significant method to unravel the problem of information overload.The core of the recommendation system centers round the recommendation algorithm.In this paper,the collaborative filtering recommendation algorithm is researched and analyzed by consulting a large number of documents.Aiming at the problems of user interest drift and data sparseness in the algorithm,this paper proposes a personalized recommendation algorithm that is based on timed forgetting model and clustering algorithm.The contents of research included in this paper are hereby demonstrated as follows:(1)This paper carries out research on the traditional user-based collaborative filtering recommendation algorithm,deeply analyzes the changes of user interest,and proposes a collaborative filtering recommendation algorithm based on timed forgetting model.First,taking the distinction of forgetting laws by user groups of different ages into consideration,this research paper establishes a time function by simulating the forgetting curve,and thus constructs the timed forgetting model after concluding forgetting factors to weight user ratings.Then,it modifies the methods of similarity calculation and rating prediction calculation.(2)Aiming at the issue of data sparseness,this paper continues to improve the recommendation accuracy on the basis of the collaborative filtering recommendation algorithm based on timed forgetting.Through researching on users' preferences for item features extracted from user rating information and item feature information,this paper builds a matrix of user preference for item category.Aiming at the problem that the FCM clustering algorithm is easy to fall into the local optimal solution,this paper also improves the FCM algorithm based on the particle swarm algorithm.Then,this paper combines the improved PSO-FCM clustering algorithm with user preference matrix to cluster users and calculates the similarity of users in the clustering.Finally,a collaborative filtering recommendation algorithm based on timed forgetting and clustering is obtained.This article compares the proposed algorithm with other existing algorithms on the Movielens dataset.Experimental results show that the proposed algorithm effectively alleviates the problems of user interest changes and data sparsity,and improves the quality of recommendation results.