Identifying top-<italic>k</italic> influential nodes in social networks: a discrete hybrid optimizer by integrating butterfly optimization algorithm with differential evolution.

The most challenge of influence maximization (IM) is to locate a finite set of influencers while maximizing the influence dissemination in a social network. Due to the increasingly widespread and complex application scenarios of IM problem, how to solve the problem effectively remains as a prominent research hotspot. However, most of the existing IM algorithms tend to prioritize either lightweight computational time or solution accuracy, which are hard to be acquired simultaneously. Therefore, considering the trade-off between efficiency and effectiveness, a novel discrete hybrid optimizer by integrating butterfly optimization algorithm (BOA) with differential evolution (DE), named DBOA-DE, is proposed in this paper. An adaptive probability is designed to guide the two operations in the hybrid optimizer, where BOA shows excellent exploratory characteristics by simulating the behavior of butterfly swarms, while DE performs local exploitation through mutation and crossover procedures. Furthermore, in expectation of enhancing the solution accuracy, an improved local search policy is conceived to avoid DBOA-DE from falling into local optimum. Extensive experiments on six real-world social networks show that DBOA-DE outperforms baseline algorithms on influence propagation while maintaining acceptable efficiency, which validate the promising effectiveness and efficiency of the proposed algorithm for IM problems. [ABSTRACT FROM AUTHOR]