An Efficient Multi-objective Community Detection Algorithm in Complex Networks

Original scientific paper Community detection in complex networks is often regarded as the problem of single-objective optimization and it is hard for single-objective optimization to identify potential community structure of meaningfulness. Thus, algorithm of multi-objective optimization is applied to the field of community detection. However, multi-objective community detection algorithm is prone to local optimization and weak diversity of the set of Paretooptimal solutions. In view of this, based on the framework of NSGAII, a multi-objective community detection algorithm, named I-NSGAII, is proposed in this paper. This algorithm is able to optimize simultaneously the two conflicting objective functions evaluating the density of intra-community connections and the sparsity of inter-community connections, and obtain the set of Pareto optimal solutions having diverse hierarchal community structures; it also proposes diversity evolutionary strategy enabling the algorithm to expand searching space and thus avoids local optimization of the set of Pareto-optimal solutions. In addition, to improve algorithm’s searching ability, I-NSGAII algorithm adopts the strategies of locus-based adjacency representation, unified label, one-way crossover and local mutation. Tests on synthetic and real-world networks and comparisons with many state-of-theart algorithms verify the validity and feasibility of I-NSGAII.