A Research And Implementation Of Improved Collaborative Filtering Algorithm Based On User Trust

With the development of the Internet,the explosive growth of information and the rapid spread of the network,search engines and personalized recommendation system has become the most popular tools for people to get information?However,when the part of the information is difficult to simply describe or the user find it difficult to accurately describe their own demand information,the search engine's role becomes difficult to play,therefore people would turn to use the personalized recommendation system for help obviously.Collaborative recommendation has a great advantage over other recommended systems.However,there are some problems that hinder the accuracy and personalization of the system,such as sparseness,cold start,and scalability.In this paper,a collaborative recommendation system based on user clustering is proposed for sparse and personalized recommendation.The main contents are as follows:1.This paper analyzes the background,significance and research progress of the personalized recommendation,analyzes the recommended process and the advantages and disadvantages of the commonly used personalized recommendation system.Finally,it points out the challenges and research hotspots of the recommended technology,and provides the theoretical basis for the further research.2.Aiming at the problem that the user similarity algorithm is a measure of user similarity,this paper introduces the user trust degree.By calculating the direct trust and indirect trust degree of the user,the comprehensive trust degree and the traditional trust The similarity degree algorithm is used to fuse,and an improved similarity algorithm is proposed to correct the nearest neighbor prediction score based on the traditional similarity degree,and the nearest neighbor of the target user is more accurate.3.In order to alleviate the problem of cold start and sparseness of cooperative filtering recommendation system,this paper introduces the user characteristic information such as age and gender,and constructs the user characteristic information matrix.Aiming at the problem that the traditional K-means algorithm clustering result is influenced by the initial center points,the maximum distance method is proposed to improve the K-means algorithm.Then,the attributes of the user characteristic information are weighted,and the weighted user characteristic information is clustered by the improved clustering algorithm.The clustering users with high similarity can be initialized in the offline stage.At the sametime,the load of the system is reduced and the system's scalability is improved.4.In this paper,several experiments are compared on the MovieLens data set,and compare with the traditional K-means user clustering collaborative filtering algorithm.The results show that the recall rate,accurate average and absolute error and rate are used to evaluate the experimental results.The recommended effect of the algorithm is better.