Measuring similarity between microblog users and its application

Analyzing user interest and building user profile is very important for microblog's user relationship analysis, which is the fundamental work for social network formation and analysis. This paper mainly discusses approaches of microblog's user relationship analysis. We view microblog's social network as a weighted undirected graph, where users are treated as nodes linked by edges, and the weights of edges mean the relationship strengths between users. This paper defines user relationship strength as user similarity, and proposes several user similarity estimation approaches by the use of various attribute information of users such as background information, tweets and social information respectively and systematically investigated them by experiments, the experimental results showed that social information-based user similarity achieved the best performance. In addition, we tested them in user recommending experiments, and social information-based user similarity also got the best recommending results. Finally we applied social information-based user similarity to generate microblog's social network, called as user similarity network, on which we conducted community mining experiments, the experimental results showed our approach is of remarkable performance.