Optimization And Research Of Recommendation System Based On Complex Network

Nowadays,with the rapid development of information dissemination technology,people are increasingly keen to browse information through various social media or maintain online social relations.In the process,the information explosion caused by the fragmentation of the audience makes it impossible for the audience to accurately capture the services that satisfy their needs.The daily complexity of information makes people's concerns very disturbed.Search engines can no longer meet the needs of most users.A recommendation system is derived in such an environment.At present,the development of the recommendation system is gradually mature.However,due to the low accuracy of personalized push and the unsatisfactory effect of recommendation,recommendation is still a field that needs continuous innovation.The core step of recommendation system is recommendation algorithm.Collaborative filtering algorithm is easy to operate and easy to understand.The algorithm mainly uses user item scoring matrix to mine near user preferences and predict the most likely items for target users.But at the same time,there are some problems such as sparse matrix,low scalability and low recommendation accuracy.Combined with the basic idea of the above recommendation algorithm and the challenges it faces,the algorithm is studied and optimized on the basis of traditional collaborative filtering.In this paper,the complex network theory is introduced into the recommendation algorithm,and a collaborative filtering algorithm based on complex network is proposed,which has been improved from a single recommendation to a variety of technologies,and the main research work is as follows:(1)A collaborative filtering algorithm based on link prediction is proposed to reduce the impact of data sparsity on the collaborative filtering algorithm.After fully understanding the problems of collaborative filtering algorithm,in order to make up the influence of sparse matrix,this paper proposes a similarity algorithm based on SimRank,which extends the improved SimRank algorithm to two points network,and uses the algorithm to calculate the similarity value between users and projects,and recommends it.(2)In view of the problem that the number of users or the large number of projects leads to the low scalability of collaborative filtering algorithm,a collaborative filtering algorithm based on overlapping community discovery is proposed.In this paper,the community discovery algorithm based on central node is studied,and an improved PageRank algorithm is proposed to discover the central node,and then the community in the network is excavated,and the most similar community is selected as the recommended group for the target user.Finally,the user is recommended by the nearest neighbor user set.The algorithm reduces computation cost and improves recommendation efficiency.Finally,a comparative experiment is conducted on the three latest data sets of Movielens to verify the accuracy and efficiency of the improved collaborative filtering algorithm.