Budgeted influence and earned benefit maximization with tags in social networks

Abstract

Influence Maximization Problem aims at identifying a limited number of highly influential users who will be working for diffusion agents to maximize the influence. In case of Budgeted Influence Maximization (BIM), the users of the network have a cost and influential user selection needs to be done within a given budget. In case of Earned Benefit Maximization (EBM) Problem, a set of target users along with their benefit value is given and the aim is to choose highly influential users within an allocated budget to maximize the earned benefit. In this paper, we study the BIM and EBM Problem under the tag-specific edge probability setting, which means instead of a single edge probability a set of probability values (each one for a specific context e.g., ‘games,’ ‘academics,’ etc.) per edge is given. The aim is to identify the influential tags and users for maximizing the influence and earned benefit. Considering the realistic fact that different tags have a different impact on different communities of a social network, we propose two solution methodologies and one pruning technique. A detailed analysis of all the solution approaches has been done. An extensive set of experiments have been carried out with three benchmark datasets. From the experiments, we observe that the proposed solution approaches outperform baseline methods (e.g., random node-random tag, high-degree node–high-frequency tag, high-degree node–high-frequency tag with community). For the tag-based BIM Problem the improvement is upto $8\%$ in terms of number of influenced nodes and for the tag-based EBM Problem the improvement is upto $15\%$ in terms of earned benefit.