rLinkTopic: A probabilistic model for discovering regional LinkTopic communities

Although geographic and regional aspects of communities find many practical applications, e.g., in social studies and marketing, to date, existing approaches to community detection have paid little attention to these features when analyzing social network data. To address these shortcomings, we introduce the concept of regional LinkTopic communities and propose a novel probabilistic model for extracting such communities. Our model jointly considers the spatio-temporal proximity of users in terms of the messages they post over time, together with contextual links and message topics to determine communities. The model allows users to have a membership in more than just one community, an important feature when discovering communities based on topics. Each community derived by our approach is not only described by a mixture of topics but also by its regional properties. Using data from Twitter, we demonstrate the effectiveness of our model in extracting regional LinkTopic communities, which are described in terms of both geographic locations and coherent topics. The experimental results show that our model outperforms related models that only use links and topics to extract communities by the measure of perplexity.