Mobile Social Network Information Filtering And Recommendation System Research

With the rapid development of mobile Internet, mobile social network has become an indispensible part of our life. Mobile social network has a large user community and the information dissemination characteristics of fast and open type, which are highly attractive for commodity marketing. But the existing mobile social network lacks a formal and friendly products publicity mechanism and an information-discriminating mechanism of fake and inferior commodities. So that various advertisings are all-pervasive and these advertidings are difficult to distinguish, which seriously affect the normal experience of the users.On the one hand, the paper researches on related available technology for spam filtering in order to solve the problem that users are affected by spam. As a result, the filtering framework based on machine learning classification method especially SVM shows high accuracy and low cost. However, SVM needs long time for training and will be not flexible to cope with the change of datasets. Focused on the weakness of SVM, an improved SVM increment learning algorithm is proposed which can save training time and to keep and higher the accuracy. The proposed algorithm is applied in a spam filtering system and function well.On the other hand, by studying the existing recommendation algorithm in the social network, especially the positive significance of social network information for goods recommendation, a hybrid recommendation system based on social network is proposed. This recommendation system excavates N users with the highest similarity in interest based on the user’s social relations network. At the same time, considering the potential demand for goods of users may be contained in the social network messages, the messages of the user are excavated and information of excavated is applied to recommend, which obtains good recommendation effects and user satisfaction.Finally, the implementation and effect of the mobile social network information filtering and recommendation system are shown in the “YouXin” system. And the filtering algorithm and recommendation algorithm also are simulational analyzed. The test results show that the spam filtering system with the improved SVM incremental learning algorithm proposed in the paper achieves good filtering effect. The hybrid recommendation method based on the social network also obtains good recommended effect on the basis of fitting subject requirements.