A novel overlapping community mining algorithm for micro-blog platform

Zhang Zhaoyin

COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(12C) 1259-1268

School of Computer Science and Technology in Heilongjiang University 150010, China

This paper concentrates on the problem of overlapping community mining for micro-blog platform, which is an important problem in social network mining. Firstly, we convert of overlapping community mining to a weighted graph computation problem, in which nodes represent users and vertex denotes the relationship between users. Secondly, we introduce the concept of user influence to solve the problem of overlapping community mining, which is a main innovation point in this paper. To calculate the user influence in the micro-blog platform, two types of micro-blog information are utilized (that is, user properties and micro-blog properties), and then the analytic hierarchy process is used to calculate the weight of each influencing factor. Furthermore, user properties contain user ID, user type, attention number, number of fans, number of micro-blog, number of mentions and so on. On the other hand, micro-blog attributes contain micro-blog number, publishing date and time, forwarding number, comment number and so on. Thirdly, a weighted network based overlapping community mining algorithm is proposed, in which the original overlapping communities are discovered in advance, and then final results are obtained by expanding the original ones. Finally, to testify the effectiveness of the proposed, experiments are conducted on several datasets and compared with other related works. Experimental results demonstrate that the proposed algorithm can detect overlapping communities in micro-blog platform with high accuracy, and our algorithm is suitable to be modified to run in the parallel mode, hence, the large-scale overlapping communities can also be solved by this proposed algorithm.