Using Maximum Weighted Cliques in the Detection of Sub-communities’ Behaviors in OSNs

Abstract

The extraction of clique relations in social networks can show information related to groups and how those groups are formed or interact with each other. In one pending application, we are showing how such “detection of abnormal cliques’ behaviors” can be possibly used to detect sub-communities’ behaviors based on information from Online Social Networks (OSNs). In social networks, a clique represents a group of people where every member in the clique is a friend with every other member in the clique. Those cliques can have containment relation with each other where large cliques can contain small size cliques. This is why most algorithms in this scope focus on finding the maximum clique. In our approach, we evaluated adding the weight factor to clique algorithm to show more insights about the level of involvement of users in the clique. In this regard clique activities are not like those in group discussions where an activity is posted by one user and is visible by all others. Our algorithm calculates the overall weight of the clique based on individual edges. Users post frequent activities. Their clique members, just like other friends, may or may not interact with all those activities.