Community Mining in Complex Network Based on Parallel Genetic Algorithm

Abstract

Community mining has been the focus of many recent efforts on complex networks, and the genetic algorithm with low time-complexity is widely used in this discipline. To enhance the performance of genetic algorithm for community detection, the modified crossover operators which are more suitable for community detection is proposed in this paper, and the heuristic mutation operator based on local modularity is designed to avoid the blindness of random flip. Additionally, to avoid premature, an independent evolution model is implemented on Chain Map Reduce framework. The experimental results show that the distributed evolutionary model contributes to reduce the selection pressure and maintains the population's diversity. Moreover, the modified genetic operators improve the global optimization performance and quicken the convergence speed.