LGP-DS: A novel algorithm for identifying influential nodes in complex networks based on multi-dimensional evidence fusion

Abstract Identifying influential nodes in complex networks is crucial in various application scenarios, such as blocking the spread of rumors, containing disease transmission and facilitating precise targeting of product advertisements. Existing researches have presented various centrality measures to identify influential nodes in the network, but most measures evaluate the importance of nodes from limited dimensions. To fill this gap, we propose a novel algorithm called Local-Global-Position based on the Dempster-Shafer evidence theory (LGP-DS) to solve the problem of identifying influential nodes. The proposed LGP-DS algorithm first calculates the information about propagation capability of nodes based on the global, local and position attributes, and thus obtains multiple evidence dimensions. Next, information entropy is employed to assess the contribution of different evidence dimensions, and the information is aggregated using Dempster-Shafer evidence theory, which facilitates the evaluation of nodal importance within a network. The effectiveness of the LGP-DS algorithm is validated by several simulated experiments on real-world networks. The results demonstrate that the proposed algorithm outperforms eight widely used algorithms in terms of discrimination power, top-10 nodes, and ranking accuracy.