An improved cuckoo search optimization algorithm with genetic algorithm for community detection in complex networks

Abstract

This paper improved Cuckoo Search Optimization (CSO) algorithm with a Genetic Algorithm (GA) for community detection in complex networks. CSO algorithm has problems such as premature convergence, delayed convergence, and getting trapped in the local trap. GA has been quite successful in terms of community detection in complex networks to increase exploration and exploitation. GA operators have been used dynamically in order to increase the speed and accuracy of the CSO. The number of populations is dynamically adjusted based on the amount of exploration and exploitation. Modularity objective function (Q) and Normalized Mutual Information (NMI) is used as an optimization function. It was carried out on six types of real complex networks. The proposed algorithm was tested with GA, Artificial Bee Colony (ABC), Grey Wolf Optimizer (GWO), and CSO, with different iterations in modularity and NMI criteria. The results show that in most comparisons, the proposed algorithm has been more successful than the basic comparative algorithms, and it has proven its superiority in terms of modularity and NMI. The proposed algorithm performed an average of 54% better in modularity and 88% in NMI than other algorithms. It performed on average in modularity criteria 84.3%, 58.8%, 33.7% and 38.8%, respectively, compared to CSO, ABS, GWO and GA algorithms, and in terms of NMI index, 188.7%, 39.1%, 52.3% and 73.8%, respectively in CSO, ABS, GWO and GA algorithms performed better.