Research And Implementation Of Microblog Friends’ Recommender Mechanism Based On Data Mining Technique

With the rapid development of the microblogging services, users can send and receive their interested contents by means of this emerging platform according to their own wishes. The microblog has become the most important social media in which its users can communicate with each other as well as disseminate information. When facing with the mass of information generated from the microblogging platform every day, how to find the useful information has become a critical issue that has drawn many concerns. As the information of microblogs is always accompanied with the microblog publishers, finding similar microblog users has become an important channel for users to get the information needed. All above make microblog friends’ recommendation become a new hotspot problem for both commercial and research communities.Based on microblog analysis and data mining technology, this thesis studies on the mechanism of microblog friends’ recommendation, which includes the construction the model of microblog friends’ recommendation and the implementation of related algorithms.Firstly, the existing microblog friends’ recommendation algorithms have been deeply studied. Related theories have been summarized, and multiple features of microblog users have been analyzed and combined. A personalized friends’ recommendation model that can utilize multiple features of microblog users including current friendships, personalized tags, location information and the hottest topics has been constructed.Secondly, based on the model constructed, a microblog friends’ recommendation algorithm is designed and implemented. Specially, a friends’ recommendation process is designed at first. Then, the process of candidate friends’ generation and the similarity calculation of user tags, location information, and interested hot topics are designed respectively. Finally, a personalized friends’ recommendation algorithm combining multiple features of microblog user is proposed.Lastly, in order to demonstrate the feasibility and advantages of the proposed model and algorithms, the cross-validation method is employed to verify the effectiveness of the proposed personalized microblog friends’ recommendation. After comparing with several common microblog friends’ recommendation algorithms in terms of Precision, Recall and F-Measure, the personalized microblog friends’recommendation algorithm proposed in this thesis has been proved effective and better than all the baseline algorithms.