Computational experiments based on competitive influence diffusion model

Abstract

Understanding the strategies to optimize/suppress information spreads under intense competition could provide important insights in a broad range of settings including viral marketing, emergency response and information system design. However, most of existing studies about competitive influence diffusion mainly focus on two-information competition mechanism. To date, the competitive influence maximization problem considering the mechanism of multi-information competition is still not well studied. In this paper, we conducted computational experiments to study the competitive influence maximization with multi-information competition mechanism. By applying an information diffusion model called limited attention model (LAM), we carried on two computational experiments to validate the model and investigate the relation between seed selection methods and the properties of information cascades. Our experimental results show that 1) the LAM model could reproduce the features of empirical distribution in Chinese social media; 2) the eigenvector centrality-based heuristic is a reasonable seed selection method for competitive influence maximization problem. The results of this paper can provide significant potential implications for information system design and management.