Improving the Performance of an Integer Linear Programming Community Detection Algorithm Through Clique Filtering

Abstract

Different fields of science use network representation as a framework to model their systems. The analysis of network structure can give us essential information about the system. However, the size of such a network can limit the applicability of some fundamental techniques like mathematical programming. Thus, here we propose a novel network size reduction technique based on a clique filtering approach. Our goal is twofold: (1) reduce the network size and speed up the community detection process, and (2) preserve the modularity of the original partition in the context of the exact model. Conducted experiments show the feasibility and correctness of the proposed technique.