Evaluating the Igraph Community Detection Algorithms on Different Real Networks

Abstract

Complex networks are an essential tool in machine learning and data mining. The underlying information can help understand the system and reveal new information. Community is sub-groups in networks that are densely connected. This community can help us reveal a lot of information. The community detection problem is a method to find communities in the network. The igraph library is used by many researchers due to the utilization of various community detection algorithms implemented in both Python and R language. The algorithms are implemented using various methods showing various performance results. We have evaluated the community detection algorithm and ranked it based on its performance in different scenarios and various performance metrics. The results show that the Multi-level, Leiden community detection algorithm, and Walk trap got the highest performance compared to spin glass and leading eigenvector algorithms. The findings based on these algorithms help researchers to choose algorithms from the igraph library according to their requirements.