Study On The Social Recommender Algorithm Combined With The User's Global Influence

The rapid expansion of the Internet brings people into the Big Data era.The huge amount of data could make users have more choices,however,it raises difficulties,a major one being the Information Overload.Because it's harder for users to find the information that they need rapidly.The recommender system can solve this problem partly,but it still exists some issues such as data sparse and cold start.Luckily,the integration of social relation information brings new vitality to the recommender system.The researchers discovered that the participation of the social information could alleviate the cold start to a certain extent.At present,most researches integrated the local influence of users into the social recommender system when they use the social relation information,but they did not paid attention to the global influence of the whole users.At present,the social recommendation algorithm is not yet be mature,there are some problems need to be solved.Combining with the research status in this field,we discuss the impact of users in social recommendation algorithm and make the global influence of the user a priority.After that,we need think about how to make full use of social relation information in the social recommender algorithm.Based on this research,this thesis constructs a new algorithm which combined users' global influence with the basic social recommender algorithm.Subsequently,we designed the experiment to prove the algorithm's feasibility.Finally,on this basis,the thesis designed a prototype system as well.The main work of this thesis is as follows:(1)State the background and history of the recommender system,analyze the status and the main problems of social recommender system.On this basis,expound the emphasis and the significance of this thesis.(2)Research the related technology of recommender system and social recommender algorithm.Expound the evaluation index of the centrality node in graph structure.Based on these technique,explore what impact could users' global influence have by doing experiments.To verify the importance of users' global influence,combine the degree centrality with the social recommender algorithm which is based on the matrix factorization.Evaluate the performance of the new algorithm by the comparing experimental on the real dataset.(3)Based on the above study,to improve the algorithm,calculate the global influence of the users in social relation network by using the modified LeaderRank and use it to adjust the users' influence in the social recommender algorithm.(4)According to the study on the social recommender algorithm combined with users' global influence,design and implementation a social recommendation system prototype,and describe the overall framework of the system and the main function modules in detail.