LAUSR

Detecting cross-communities constructed based on the commonalities of their adaptive social traits is crucial for solving many settings of real-world problems, such as examining the dynamics of a social network, examining the behavior patterns that influence the outbreak and spread of disease, determining a criminal organization’s influential individuals, and amplifying a business potential. Unfortunately, investigating approaches that emphasize the detection of such cross-communities has been understudied. Moreover, few current approaches may detect cross-communities but without regard for their granularities. To overcome this, we introduce a novel methodology that analyses the overlapping among social traits to detect the most granular cross-communities with multisocial traits. The methodology is implemented in a working system called CDBCD. It can detect the most granular multisocial traits cross-communities, to which an active user (i.e., a context user) belongs. Such cross-communities are detected and selected by the system if they exhibit evidence of multisocial traits’ homophily among their users. We propose novel context-driven search techniques that infer the relationships among various social traits. We evaluated CDBCD by comparing it with ten methods. The results demonstrated that CDBCD can detect granular cross-communities with marked accuracy. The improvement of CDBCD over the ten methods combined is 23% and 19% in terms of ARI and F1-score, respectively.