Influence contribution ratio estimation in social networks

Abstract

Social networks are becoming an ideal choice for marketing activities and advertising campaigns due to the explosive growth of social network users. In these advertising campaigns, influential users, such as celebrities, often post or repost the ads to help disseminate product information through the ‘word-of-mouth’ effect in social networks. It is significant to allocate remuneration fairly to these influential users based on their contributions to disseminating ads. To address this, we propose a concept called the influence contribution ratio, which represents the contribution ratio of each influential user to an advertising campaign. We introduce two types of coalitional games to depict the process of influence diffusion from multiple levels, namely macro and micro coalitional games, and propose a metric called InfConR based on the Shapley value in coalitional game theory to measure the influence contribution ratio fairly. A naive method to calculate the InfConR value for each user is to use a Monte Carlo (MC) simulation to enumerate a certain number of cascades for the advertising campaign. However, this method is too time-consuming and not realistic. Therefore, we propose a scheme called ICR, which involves two components: 1) sampling algorithms for InfConR in the Independent Cascade (IC) model and Liner Threshold (LT) model, respectively, and 2) an algorithm with approximation guarantees to minimize the sampling number. Our experiments on four real-world datasets demonstrate the superiority and effectiveness of our scheme.