Research On Collaborative Filtering Recommendation Algorithm Based On User Clustering

With the popularity of the internet,a lot of meaningless data makes it difficult to filter effective information.To help people quickly and effectively filter information,the personalized recommendation systems were proposed.The recommended algorithm as the core of the recommendation system has always been the focus of the study.The collaborative filtering algorithm is the most widely used in many of the recommendation algorithms.It discovers the user's preference by mining the user's historical behavior data,groups the users based on different preferences,and then recommends the items.However,with the increasing number of users and items in e-commerce system,the data sparsity and recommended efficiency gradually become the bottleneck that restricts the development of collaborative filtering algorithm.In order to improve the recommendation quality and efficiency of the collaborative filtering algorithm,this thesis proposes an improved collaborative filtering recommendation algorithm based on user clustering.And then designs and implements a movie recommendation system of B/S architecture based on the improved algorithm.This thesis introduces the development background and architecture design of the personalized recommendation system,and gives the basic idea and the main problems of the traditional collaborative filtering algorithm.And then improves the traditional algorithm from the two aspects of the offline user clustering and user similarity calculation.The improvement of the algorithm is mainly reflected in two aspects.First,considering the influence of user rating information and item category preference information on user clustering,and proposes a joint user clustering algorithm.The algorithm is based on user rating information and item category preference information to cluster the basic users,generating two clustering centers and two user categories matrix.Then calculate the similarity between the target user and cluster center,to find the cluster of the target user.The two clusters are merged and subtracted to get the nearest neighbor search space of the target user.Secondly,a weighted Pearson correlation coefficient calculation method based on the difference factor is proposed to solve the problem that the similarity of traditional Pearson correlation coefficient is not sensitive to absolute value.Using difference factor as the weight to adjust the traditional Pearson correlation coefficient.Using MovieLens data set,the values of MAE,accuracy and recall rate and F1 value as metrics,through many experiments to evaluate recommended effect of improved algorithm,traditional user-based collaborative filtering algorithm(CF)and traditional collaborative filtering algorithm based on user clustering(UCCF)algorithm,the experimental results show that the improved algorithm can effectively improve recommended efficiency and recommended accuracy.Finally,designed and implemented a movie recommendation system based on improved algorithm.The system adopts douban Top250 movie information as the data set,using PHP and Matlab mixed programming to provide personalized recommendations for users based on user preferences.