Research On Recommendation Algorithm Based On User Multidimensional Social Network

With the rapid development of the Internet industry,people's life has been inseparable from the Internet a lot,when enjoying the benefits of popular Internet,information overload and other issues are becoming increasingly serious.When finding products fastly and accurately gradually become a luxury,the personalized recommendation system came into being,it can interact with the user in the process,by studying user's history browsing records or other information,access to its Interest preferences,search users' interested goods from massive information and make recommendation to users,so as to meet the user information on the diversification and personalized needs.The emergence of recommended systems,can be a good solution to the problems faced by the Internet.Recommended system based on user behavior,preferences and other data,mining the user's interest,so as to recommend valuable or interested information for the user.And it can effectively dig out the long tail information,so that valuable information can be effectively spread and be concerned,with a high degree of commercial value.In practical applications,users are always in a wide range of social relationships,and one-dimensional network analysis is tailored to specific areas of the problem.For personalized recommendations,the need to analyze the user's interest is reflected through the user's various network activities,the recommended system's task is to help users find previously found new content,research multi-dimensional overlapping network to help users in more Kind of behavior under the behavior and interest preferences,so that all-round better description of the user's interest,to make the personalized recommendation more accurate.Based on the analysis of the multi-dimensional user 's historical behavior,the algorithm uses the CPM clustering method which can identify the overlapping network cluster to find the neighbor user,synthesize the nearest neighbor of the multi-dimension,and give the recommended item when the recommended list is given Gini coefficient,to reduce the frequency of popular goods appear in the list,to avoid the popular goods become more popular,popular products better popular,increase the recommended list of diversity.Then,the recommendation strategy of this paper is introduced and the validity of the algorithm and the recommended quality are verified from different angles.The experimental results are given in the essay.The final recommendation algorithm can accurately calculate the similarity between users and meet the needs of the recommended list of diversity,so that the recommended quality to further enhance the recommended results more personalized and humane.