Community Detection And Evolution Analysis Based On Microblogging Platform

Nowadays, dozens of domestic and international well-known social networking sites, such as Facebook, Google+, Twitter, QQ zone, Sina weibo and Renren, these social network platform or application plays an important role in people’s daily life.Community structure is an important characteristic in social network. How to find these behaviors has an important significance on community structure analysis and change forecast in the social network. Community evolution is the basic characteristics of complex network, and is the result of interactions between the network structures itself.With the data with time attribute in social network accumulated, the evolution of the research community has gradually become a hot spot.Weibo is a brief real-time platform where the information are shared and disseminated by following and followed mechanism. As an important part of social networking sites, weibo become the main channel for Internet users to access information and communication.However, existing universal community detection algorithm can’t applied to the weibo platform environment. Based on the classic MetaFac tensor joint decomposition algorithm, we put forward an algorithm which combines weibo contents, such as multiple behavior data, and improve the algorithm of user interest mining.Improvements include the following three points: according to the characteristics of weibo environment, a generalization of weibo social networking element graph model is proposed; based on the MetaFac framework, for a number of parameters which contribute to the quality of community structure, a measure determining the number of effective communities is proposed; according to the measure of a variety of heterogeneous relation to the community, a multi relational weight distribution mechanism is proposed. This paper proposed the main idea of the algorithm, namely,tensor combined decomposition algorithm to combine multiple user behavior and weibo content relation of relational data community detection, the similar behavior or focus on the theme of the user belongs to the same community. Additionally, we put forward a mechanism which combined with the community detection framework to find the changes of community structure in microblogging social network over time. This paper uses the micro-blog Sina dataset to verify the performance of the improved algorithm,the experimental results show that the analysis of micro-blog content does enhance thedegree of association between the users.In this paper, we combined tensor decomposition algorithm for community evolution discovery mechanism, as much as possible from the continuous time fragments found community structure to keep a certain degree of similarity in order to ensure a smooth transition of the evolution of the community structure. Based on the theory of this mechanism, the data of continuous time segment is found in the incremental community, and the changes of community structure over time are analyzed from the aspects of users, themes and other important entities.This paper designs a data acquisition system of weibo based on open source framework, the system of adaptive extension of Scrapy crawling strategy module and database access etc. We crawled a large number of micro-blog Sina data for analysis of the characteristics of micro-blog social networking sites and the proposed performance evaluation algorithm.