LPC: A Local Path-Based Centrality Method for Identifying Influential Nodes in Temporal Networks

In network analysis, centrality methods are essential for identifying the “most influential” nodes within complex networks. However, many real-world networks are temporal, with time-dependent edges or node interactions. Many centrality methods have been proposed in the literature to find the most influential nodes. Among them, temporal betweenness, temporal closeness, temporal PageRank, and temporal degree are popular methods to capture the node’s importance. The temporal betweenness and temporal closeness are computationally demanding and require extensive global information. Temporal PageRank struggles with convergence issues. Although temporal degree centrality is more straightforward and computationally less intensive, it fails to fully capture influence dynamics, mainly when significant neighboring nodes are absent. To address this problem, we propose a novel method called Local Path-based Centrality (LPC). LPC evaluates node importance by considering local paths and timestamps associated with each node. Experimental results on eight real-world temporal networks demonstrate that LPC accurately identifies the most influential nodes. The method’s effectiveness is validated using three performance metrics: the Kendall τ correlation coefficient, improvement percentage, and the network robustness value. LPC shows a strong correlation with the Susceptible-Infected-Recovered (SIR) epidemic spreading model and exhibits the lowest robustness values compared to other established centrality methods.