Solving the robust influence maximization problem on multi-layer networks via a Memetic algorithm

Abstract

The influence maximization problem is aimed at determining influential nodes as seeds to reach the maximal influential range. Considering the wide application in marketing and social dynamics, increasing attention has been paid to modeling the information diffusion process and efficient seed selection algorithms on both single-layer and multi-layer networks. Interestingly, some recent studies indicate that the robustness of seeds in the diffusion process against potential disturbances like structural failures is significant in applications. But the current study only considers scenarios on single-layer networks. Meanwhile, multi-layer networks have shown non-negligible values in theoretical analyses and practical applications; the study on the robust influence maximization on such networks is urgent but remains to be an open question. Therefore, this paper gives the design of a performance measure to evaluate the influence ability of seeds on multi-layer networks under structural destructions. And a rational configuration of the included changeable parameter is determined based on qualitative analyses. Further, a Memetic algorithm, termed MA-RIMMulti, has been devised with several problem-orientated operators to find influential and robust seeds. This algorithm successfully solves the robust influence maximization problem on multi-layer networks, and shows competitive performance over existing optimization methods on several synthetic and real-world networks. Additionally, the effect of structural changes on the performance of seeds is also studied, and the topological rewiring technique is validated to be effective to improve the influence range of seeds. From the perspective of the influence maximization problem and its robustness, the underlying information behind multi-layer networks has been excavated. Meanwhile, key nodes in such networks can be found via the proposed approach to facilitate possible tasks in social propagation and information diffusion.