A hybrid evolutionary algorithm for community detection

Abstract

Evolutionary algorithm belongs to the behaviorism which is one of major approaches to artificial intelligence. Community detection is one of the important applications of the evolutionary algorithm. Detecting the community structure, an essential property for complex networks, can help us understand the inherent functions of real systems. It has been proved that genetic algorithm (GA) is feasible for community detection, and yet existing GA-based community detection algorithms still need improving in terms of their robustness and accuracy. A Physarum-based network model (PNM) with an intelligence of recognizing inter-community edges based on a kind of multi-headed slime mold, has been proposed in the phase of GA's initialization for optimization. In this paper, integrated with PNM after three operators of GA during the process of community detection, a novel genetic algorithm, called P-GACD, is proposed to improve the efficiency of GA for community detection. In addition, some experiments are implemented in five real-world networks to evaluate the performance of P-GACD. The results reveal that P-GACD shows an advantage in terms of the robustness and accuracy, contrasted with the existing algorithms.