2025/03/21 by Megh Singhal, Bhawna Saxena · 1 voice
Physics and Astronomy · Psychology · #Complex Network Analysis Techniques #Mental Health Research Topics #Opinion Dynamics and Social Influence
paper · doi:10.1142/s021952592540003x
openalex publication_date 2025/03/21 · openalex created_date 2025/10/10 · openalex updated_date 2026/06/26
Diffusion models drive the simulation for computing the expected spread of seed nodes, thereby helping in assessing the goodness of the chosen seeds. Without the inclusion of a model for simulating diffusion, the process of influence maximization (IM) remains incomplete. Given the co-existence of positive and negative relationship in real-world social networks, it becomes necessary to account for negative influences, as they can reverse the direction of influence diffusion. Additionally, the strength of relationships varies among social actors, with some links being strong and others weak. This variability in relationship strength significantly influences a node’s probability of sharing information with its neighboring nodes. Inspired by these actualities, we propose a novel Edge Sign and Strength-based Influence Diffusion (ESS-ID) model for influence diffusion in signed-weighted social networks. It calculates the likelihood of influence propagation from one node to another based on the weights of the edges. The sign associated with an edge is utilized to compute the actual positive influence exerted by a node. The performance of ESS-ID model is compared with three prevalent models used for studying diffusion in signed social networks, with emphasis not only the attainment of a large influence spread but also the achievement of a substantial positive influence spread. The comparative study was performed on 3 real-world signed social networks, and the obtained results establish that proposed ESS-ID model outperforms the other models by huge margins. ESS-ID is found to have attained an influence spread of approximately 85–90%, which is much higher compared to the other widely used models.