Inferring the densest multi-profiled cross-community for a user

Abstract

The most realistic and widely formed social structures are granular multi-profiled cross-communities constructed from users sharing the traits of common adaptive multi-social profiles. The more important types of such cross-communities that deserve attention are the densest holonic ones with various adaptive multi-social profiles, because they exhibit many interesting properties. The likelihood of an exact match between an active user and his/her cross-community’s interests increases as the cross-community becomes denser. Moreover, the denser such a cross-community is, the more distinct is its interests and the more distinguishable these interests from the ones of other cross communities. Unfortunately, methodologies that emphasize the detection of such granular multi-profiled cross-communities have been understudied. To overcome this, this study proposes a novel methodology that analyzes hierarchically overlapped social profiles to detect the smallest and the most granular multi-profiled cross-communities. The methodology is implemented in a working system named OMACexplorer. The system can detect the densest multi-profiled cross-communities from heterogeneous information and social networks. It can also infer an active user’s densest multi-profiled cross-community that matches his/her own social traits. We evaluated OMACexplorer by comparing it experimentally with eight well-referenced methods. Based on the experiment results, the improvement of OMACexplorer over the other eight methods are 47% and 51% in terms of ARI and F1-score, respectively, which demonstrate the high efficiency and effectiveness of the proposed method.