Distributed community detection in social networks with genetic algorithms

Community detection in social networks is a hot research topic that has received great interest in the recent years due to its wide applicability. This paper proposes a scalable approach for community structure identification using a genetic algorithm. Two existing fitness functions are analyzed and genetic parameters are tuned on thoroughly studied networks with known community structures. Experiments on a large data set show how the amount of time necessary to determine meaningful communities in a network is significantly reduced by running the algorithm distributed. This enables the analysis of larger, real-world networks. We then propose a new fitness function that offers a good tradeoff between efficiency and speed.