Parallelizing SLPA for Scalable Overlapping Community Detection

In a very general context, communities in networks are defined as groups of nodes that have some common properties such that connections are stronger between the nodes in a community than with the nodes in the rest of the network. It is quite common for nodes to participate in multiple communities. Therefore a community detection algorithm for such applications should be able to detect overlapping communities. However, overlapping community detection is more computationally intensive than disjoint community detection and presents new challenges that algorithm designers have to face. Besides, the big data phenomenon with exabytes of data brings up datasets that take too long to analyze using even the fastest algorithms currently available. Fortunately, the amount of computing power available to researches also increases. This computing power usually comes structured as a number of cores, processors, or machines joined together to form a high performance computer, cluster or a supercomputer. In this paper we analyze what other researchers have done to utilize high performance computing to perform efficient community detection in social, biological, and other networks. We use the Speaker-listener Label Propagation Algorithm (SLPA) as the basis for our parallel overlapping community detection implementation. SLPA provides near linear time community detection and is well suited for parallelization. We explore the benefits of a multithreaded programming paradigm and show that it yields a significant performance gain over sequential execution in detecting overlapping communities.