Identification of top-k influential nodes based on discrete crow search algorithm optimization for influence maximization

Abstract

Influence maximization refers to selecting a small number of influential nodes in a given network to maximize the influence affected by the subset. In social network analysis and viral marketing, influence maximization is greatly significant. The greedy-based algorithm is time-consuming in estimating the expected influence diffusion of a given node set, which is unsuitable for large-scale network. The traditional heuristics often have the problem of low accuracy. In this study, in order to solve the influence maximization problem more effectively, a meta-heuristic discrete crow search algorithm (DCSA) using the intelligence of crow population is proposed. In DCSA, a new coding mechanism and discrete evolution rules are constructed. The degree-based initialization method and the random walk strategy are adopted to enhance the search ability. Moreover, according to the network topology, influential nodes Candidates are generated to avoid blindness in the process of crow search. Extensive experiments are conducted on six real-world social networks under independent cascade (IC) model, the results show that DCSA outperforms other state-of-the-art algorithms and obtains comparable influence diffusion results to CELF but with lower time complexity.