EIM: A Novel Evolutionary Influence Maximizer in Complex Networks

Abstract

This study addresses influence maximization in complex networks, aiming to identify optimal seed nodes for maximal cascades. Greedy methods, though effective, prove inefficient for large-scale social networks. This article introduces a double-chromosome evolutionary algorithm to tackle this challenge efficiently. This method introduces a smart operator for stochastic selection based on the node degree to initialize the primary solutions. A novel smart approach was also employed to improve the convergence of the proposed method by ranking the nodes existing in the current solution and using a blacklist to reduce the probability of selecting the nodes that might be influenced by the selected nodes. Moreover, a novel local search operator with appropriate efficiency was proposed to increase influence. To maintain solution diversity, a population diversity retention operator is integrated. Experimental evaluations on six real-world networks revealed the algorithm’s superiority in terms of influence rates, consistently outperforming the DPSO algorithm and ranking second to CELF with minimal margin according to statistical analysis using the Friedman test. For runtime efficiency, the proposed method demonstrated significantly shorter execution times compared to CELF and DPSO, showcasing its scalability and robustness. These results underscore the method’s effectiveness for applications requiring accurate identification of influential nodes.