Distance-aware optimization model for influential nodes identification in social networks with independent cascade diffusion

Abstract

Influence maximization (IM) is a challenge in social networks, which depends on the spreader selection. We propose a quadratic programming model to identify a fixed number of initial spreaders to affect the maximum nodes within the minimum diffusion time. We solve this model using a new Distance Aware Spreader Finding (DASF) algorithm independent of the community detection problem. On large-scale social networks, DASF selects anchor nodes by a novel threshold. Then a social distance is defined between anchor nodes via random walk processes. This distance is regularized by the neighborhood degree. Our model finds influential spreaders under the Independent Cascade (IC) diffusion model. It implicitly maximizes the local coverage of spreaders and minimizes the global overlap. We extract the solution of this bi-objective model by finding the principal eigenvector of the regularized distance matrix. Comparing DASF with nine algorithms on various large-scale social networks indicates that DASF performs well based on the influence spread and diffusion rate criteria. The robustness of DASF is also acceptable dealing with different noisy scenarios.