A Movie Recommendation System Based On Hybrid Double Clustering

The development of mobile Internet has brought great convenience to people's life.There is a large amount of information on the Internet for users to refer to and consult.However,it is also very difficult for users to quickly find the information they need from a wide variety of information,and it also reduces the utilization of resources.In order to solve these problems,a personalized recommendation system came into being.Current popular e-commercial recommendation systems implement a collaborative filtering approach for their recommendations.It mines content that users might like based on the user's preference and historical rating data to generate recommendations.However,the collaborative filtering algorithm is not perfect.It has data sparsity problems and cold start problems.Especially in the case of big data,these problems have become more prominent.Because the collaborative filtering algorithm is recommended according to the user's ratings of the items.In the case of big data,each user's rated items are limited but the number of the users and items is huge which leads to the sparsity of the data and seriously reduces the accuracy of the recommendation.What's more,there are too many new items and new users in the system,but they can't be recommended because the system doesn't have their historical information.This will lead to the cold start problem.Based on the above problems and the previous research,we continue to the optimize the algorithm.The main work of this thesis is as follows:(1)In order to solve the data sparsity problem,a collaborative filtering algorithm based on k-nearest neighbor(KNN),singular value decomposition(SVD)and double clustering is proposed(KSDC-CF).We use KNN technology to fill the user-item rating matrix by finding the missing data and using the neighbors' rating to fill it.And SVD technology can be used on matrix and can reduce huge matrix 's dimension which makes the matrix more intensive.We combine the user-based clustering with the item-based clustering method which not only reduces the running time,but also combines the advantages of item and user clustering and increases the accuracy.In order to verify the effectiveness of the proposed algorithm,we conducted experiments on the public dataset movielens.We compare it with the traditional collaborative filtering algorithms.The results show that the algorithm has certain effects on improving the recommendation accuracy and solving the sparsity.(2)In order to solve the cold start problem,we proposed a content-based and double clustering collaborative filtering algorithm(CHCF).The double clustering collaborative filtering algorithm is integrated with the content-based recommendation algorithm,and the user-item rating matrix is changed into the user-item attribute rating matrix,which improves the accuracy of the recommendation.More importantly,based on the existing algorithms,we combine the user feature matrix and the item attribute matrix for recommendation.For new users or new items,we firstly cluster their attributes to ensure they can find the nearest neighbor and get recommendations which solves the problem that new users can't get the recommended because the system doesn't have their historical information.In order to verify the effectiveness of the proposed algorithm,we conducted experiments on the public dataset movielens.We compare it with the traditional collaborative filtering algorithms.The results show that the algorithm has certain effects on improving the recommendation accuracy and solving the cold start problem.(3)Based on the algorithm research,this paper designs and implements a personalized movie recommendation system.The system is developed based on the Django framework.The main work includes requirements analysis,algorithm implementation,database design and framework design.And the paper shows the specific operational effects of the system.