HLPA: A hybrid label propagation algorithm to find communities in large-scale networks

Abstract

Fast detecting communities is challenging in large-scale real-world social networks and an important task in many scientific domains, such as Complex Networks and Social Network Analysis. In this paper, we propose a Hybrid Label Propagation Algorithm (HLPA) for finding communities on large-scale real-world social networks. And we conduct our experiments on real-world social networks datasets and get meaningful community results. Our method can get detection results on large-scale networks significantly fast, due to the following two benefits. The first is that our near linear algorithm HLPA is using a novel hybrid updating scheme, label decaying strategy and different initialization methods on different networks to improve the quality and scalability for detecting communities. And the second is that this is the first attempt implementation at community detection on the lightning-fast cluster computing framework Dpark, which is a Python version of Spark. Through experiment, we compare our algorithm with the state-of-art algorithms, and have confirmed our algorithms' superiority and universality for working on unweighted overlapping community detection of large-scale real-world social networks.