Influence maximization in multilayer networks via improved Coulomb algorithm based on layered neighborhood centrality

Identifying a set of key individuals that can maximize the propagation of information in multilayer networks remains a fundamental issue in influence maximization. However, existing research on the influence maximization problem in multilayer networks exhibits deficiencies in simulating dynamic paths and propagation dynamics of cross-layer information interactions, while lacking effective methods for mining key nodes in multilayer structures. To address such challenges, a novel dual-factor multilayer independent cascade (DF-MLIC) model is proposed in this paper. This model integrates global inter-layer structural differences with local node-level propagation potential, derived from clustering coefficients, to dynamically determine cross-layer propagation probabilities across layers and nodes. Furthermore, an improved Coulomb algorithm based on layered neighborhood centrality (LNCC) is proposed to effectively solve the influence maximization problem in multilayer networks. The algorithm defines a layered neighborhood centrality by introducing the variation of the Fiedler value, which unifies the measurement of a node’s local influence and its inter-layer bridging role, thereby identifying key nodes capable of efficient global propagation. Experimental results on four synthetic networks and twelve real-world networks demonstrate that the proposed LNCC more effectively identifies pivotal nodes in multilayer networks, achieving an average performance improvement of 3.6% compared to state-of-the-art algorithms.