Applying Overlapping Community Detection Based on Data Field Theory For Twitter Audiences Classification

Recent years have witnessed the increasing development of Twitter. Given a specific target, all the followers and other participants who have interactions such as mentions with target, are called audiences of the target. As one of the significant applications in social media marketing, classifying Twitter audiences is inherently equivalent to potential customer classification. In this paper, we propose to employ community discovering methods as a solution. On one hand, the significance and influence range of nodes are different, and the influence can be radiated among non-connected nodes as well. To this end, we propose to introduce data field theory to measure the strength between nodes. On the other hand, to minimize parameter selection and settings, we propose a multi-step method by first determining initial clusters based on potentials, and then applying a rough clustering algorithm to detect overlapping communities. Moreover, we collect real world Twitter dataset for the evaluation of proposed method.