Influence Propagation in Social Networks: A Comparative Analysis of Diffusion Models

Understanding how influence propagates in social media is essential for improving content virility, engagement strategies, and digital communication. With platforms like Instagram playing a key role in marketing and information dissemination, accurately modelling influence spread helps predict user behaviour and optimize influencer strategies. This research compares the performance of three diffusion models—Susceptible-Infected Recovered (SIR), Heat Diffusion, and the Hawkes Process—to understand their effectiveness in simulating influence propagation. The assessment used total influenced nodes and peak influence step along with nodes influenced at peak and steps to saturation as metrics. The Hawkes Process proved to be the most effective model, influencing 209 nodes in just six steps to peak influence. The SIR model performed at a fair level by affecting 152 nodes through quick propagation, but it reached its maximum rapidly. When applied to network data, the Heat Diffusion model exhibited a methodical and enduring diffusion characteristic as it spread to 183 nodes rather than the rapid but brief propagation pattern demonstrated by the Heat Diffusion model. The results emphasize the superiority of time-sensitive models, such as the Hawkes Process, in capturing rapid or viral influence propagation more effectively than traditional approaches. This research contributes to understanding how different models simulate social media influence and provides valuable insights for influencer selection, content optimization, and marketing strategies. Future research could explore hybrid models that integrate elements of these approaches, offering improved accuracy for diverse propagation patterns and addressing the complexities of cross-platform dynamics.