Community clustering based on trust modeling weighted by user interests in online social networks

Online social networking websites provide platforms through which users can express opinions and preferences on a multitude of items and topics, and follow users and information, and flood it by retweeting. User-user interests vary, and based on the users' interests, they can be grouped to multiple implicit interest communities. However, every interaction and user may not be trustworthy. Capturing the user's interaction with others, and predicting user interest and trust from the interactions are important parts of social media analytics. In this paper, we propose community clustering for implicit community detection based on trust and interest modeling. The trust modeling is weighted by the user's interests to group the users in multiple clusters having higher interest and trust similarity within a cluster. The proposed community clustering algorithm begins by ranking the nodes by the weighted degree and then selecting the initial community centers that are not in the neighbors of each other's. We then assign the user to the community with whom the user has the higher interest and trust similarity and higher common connections topology. We provide a probabilistic trust model to predict the unknown reliable trust between users considering their friends. We model user interests based on preferences and opinions, as well as the content experienced in social media. Furthermore, we evaluate the proposed algorithm comparing publicly available datasets with well-known algorithms for clustering quality.