Design And Implementation Of Personalized Recommendation System Based On Social Network

With the rapid development of social networks,it is difficult for people to quickly obtain the desired information in massive data,so the information overload problem is gradually formed.Enterprises and scientific research institutions have carried out a lot of research and application for this purpose,and finally the recommendation system came into being.The recommendation system can provide personalized information recommendation services based on users' interests,historical behaviors and other information.It gradually plays an important role in e-commerce,teaching,government and enterprise,and becomes a hot research topicBut,with the rapid growth of social network data,the traditional recommendation algorithms have been unable to meet the user's accurate recommendation needs,especially sparse problem,cold starting till now leading to the decline of recommendation accuracy,which has not been well solved.At the same time,the large amount of information in the social network is another great burden.The accuracy of recommendation is seriously affected.It is impossible to carry out real-time personalized recommendation.In addition,traditional recommendation algorithms only consider the direct social relationship among users,but ignore the potential interest preference characteristics and community clustering characteristics of users.In view of the above situation,the research contents of this paper are as follows:(1)By combining social network community partitioning technology with collaborative filtering recommendation algorithm,a recommendation algorithm—CBRA based on improved COPRA community partitioning is proposed.Firstly,according to user similarity matrix and user trust,user relationship matrix is constructed.Then the improved COPRA community discovery algorithm is used to divide the social network into communities and form a set of users' nearest neighbors,which effectively alleviates the problem of data sparsity and cold start.Finally,use the neighbor user ratings in the community to predict and score the items that the target user has not scored and generate a TopN recommendation list.Experiments verify the accuracy and efficiency of the proposed algorithm.(2)A collaborative filtering recommendation algorithm PKM-UserCF based on user feature clustering is proposed.Firstly,the Canopy clustering algorithm and Max-Min distance algorithm are used to improve the classical clustering algorithm K-Medoids.The results of Canopy preliminary clustering are provided to K-Medoids as the initial K value selection,which avoids the uncertainty of K value selection.Then Max-Min distance algorithm is used to optimize the selection of clustering centers and effectively avoid falling into the problem of local optimal solution.For large-scale social user data,MapReduce computing framework is used to parallelize the algorithm.Secondly,combined with the user's characteristic attributes,interest preferences and similarity,the selection of the nearest neighbor set of the target user is reduced from the whole range of social network to clustering cluster,and then TopN recommendation of the target user is realized based on user collaborative filtering algorithm.Experiments verify the accuracy and efficiency of the proposed algorithm.(3)This paper designs a personalized movie recommendation system based on the proposed recommendation algorithm using Hadoop distributed framework.The system has the following functions:?.The system can recommend movies to users according to their operation history,and users can search any movies they want to see and score the movie;?.The system provides users with the promotion of popular movies and the inquiry of movie lists;?.The system has the characteristics of social network.Users can add ones with the same interests as friends,and friends can share,forward,like,comment on their own dynamics;?.The system has community functions.There are many users with similar interests in the community.Users can join the community of their own interest,and have the functions of posting,commenting and likes.