Enhanced Twitter Community Detection using Node Content and Attributes

Twitter community detection has introduced as a new technique to discover communities in social networks. Users in social networks can be clustered dependent on the type of relationships between them such as who have interacted a lot or who have Tweeted same topics. Community detection is used to split a network into groups of closely connected nodes. Most of the existing research focus on a structure network to detect communities that only used link relationships between users, such as follower/following relationship to discover the community. Therefore, community detection may have a poor quality due to Twitter is an unstructured and noisy link relationship. Our aim is to enhance community detection in Twitter by exploiting user content and attributes in social networks to improve the quality of community detection. In this paper, we propose a technique that combined user content and attributes to enhance community detection. Our approach has enhanced the quality function of modularity and a Constant Potts Model (CPM) for Leiden community detection by using the topic model and Twitter attributes (retweet and mention). The result of our experiments on a real Twitter network and benchmark network show improvement in quality of a Leiden Algorithm for both modularity and CPM.