Edge attribute-enhanced community discovery in social networks

This thesis investigates the problem of community discovery. We define and propose solutions for three sub-problems. First sub-problem is discovering temporal interaction biased communities. To solve this problem we proposed an influence propagation model and TIS-Community detection. Second sub-problem is partitioning a graph for overlapping community discovery. Our solution involves developing an objective function to partition the graph by decomposing data and distributing them evenly across the available processors. Third sub-problem is discovering positive-persistent communities. We developed two models to address this problem. We evaluate all proposed solutions over real social networks and compare our solutions with state-of-the-art methods.