Study On Hybrid-Recommended Techniques For Papers Based On Community Discovery And Association Rules

Related researches have shown that the proportion of highly educated netizen in the usage of social networking sites is showing a declining trend, and the main reason is that the social networking sites a are a waste of time and also a lack of professionalism. In order to meet the needs of highly educated netizen, the number of academic social networking sites (social networking sites which are to conduct academic exchanges established) is in rapid increases. However, with the rapid development of the Internet, the number of scientific papers are in order of 6%-8% annual growth rate, the dramatic increase in the amount of information of scientific papers has greatly increased the difficulty of retrieving the paper, which led to the passive services in paper sharing, that is, users can not take the initiative to quickly retrieve interesting scientific papers.In order to solve the problem of passive services in sharing academic scientific papers in social networking site, we propose and implement a hybrid algorithm which combine the feature of social networking sites with traditional personalized recommendation system in order to provide the scientific papers which the user is interested for the user. Firstly, this article researches into the related technologies about traditional personalized recommendation algorithm, including content-based filtering recommendation algorithm, collaborative-filtering algorithm, association rules algorithm and hybrid recommendation algorithm. Then study the social network-based personalized recommendation technologies, including those based on graph theory algorithms and hierarchical clustering algorithms. On this basis, analyze the shortcomings of traditional personalized recommendation algorithm, and study the feasibility of combing the community discovery algorithm with the traditional personalized recommendation algorithm. Then put forward a hybrid recommendation algorithm which combines the classical community discovery algorithm---GN algorithm with association rule algorithm---Apriori algorithm. The hybrid recommendation algorithm mainly includes three parts:the first use of GN algorithm to construct user groups with similar interest, followed by the use of Apriori algorithm for association rule mining project to recommend papers, and last the recommending of paper which introduces a degree of paper’s interest. Based on proposing the hybrid recommendation algorithm, then analyze it with absolute performance and relative performance.In absolute performance, according to the correct rate and the recall rate, two criteria evaluation, analyze the results of the paper which is recommended. In relative performance, analyze the recommendation based on Apriori algorithm with the recommendation based on hybrid algorithm. Through experimental results, hybrid recommendation algorithm based on association rules and community discovery can effectively improve the quality of the paper recommended.In this article, after verifying the validity of the hybrid recommendation algorithm, design and implement a thesis personalized recommendation system. This system can push interesting papers to the users, thus saving the user time to retrieve the papers, and get a better user examination.