Efficient Algorithms for Social Influence Problems with Large Networks

Abstract

In recent years, the dizzying explosion of data and information results from social networks with millions to billions of users, such as Facebook, YouTube, Twitter, and LinkedIn. Users can use online social networks (OSNs) to quickly trade information, communicate with other users, and keep their information up-to-date. The challenge of spreading information on social networks that arises in practice requires effective information management solutions, such as disseminating useful information, maximizing the influence of information transmission, and preventing disinformation, rumors, and viruses from being disseminated. Motivated by the above issues, we investigate the problem of information diffusion on OSNs. We study this problem based on two models, Independent Cascade (IC) and Linear Threshold (LT), and classical Influence Maximization (IM) in online social networks. In addition, we investigate various aspects of IM problems, such as budget variations, topics of interest, multiple competitors, and others. Moreover, we also investigate and apply the theory of combinatorial optimization problems to solve one of the current concerns in social networks, maximizing the influence on the groups and topics in social networks. In general, the main goals of the Ph.D thesis proposal are as follows. 1. We investigate the Multi-Threshold problem for IM, which is a variant of the IM problem with threshold constraints. We propose an efficient algorithm that IM for multiple thresholds in the social network. In particular, we develop a novel algorithmic framework that can use the solution to a smaller threshold to find that of larger ones. 2. We study the Group Influence Maximization problem and introduce an efficient group influence maximization algorithm with more advantages than each node’s influence in networks, using a novel sampling technique to estimate the epsilon group function. We also devised an approximation algorithm to estimate multiple candidate solutions with theoretical guarantee. 3. We investigate an approach for Influence Maximization problem with k-topic under constraints in social network. More specifically, we also study a streaming algorithm that combines an optimization algorithm to improve the approximation algorithm and theoretical guarantee in terms of solution quality and running time.