Pareto Optimization for Influence Maximization in Social Networks

Kai Wu, Jing Liu, Chao Wang, Kaixin Yuan

Abstract: Influence maximization is the problem of finding a small subset of seed nodes to maximize the spread of influence in a social network and is an NP-hard problem. In this paper, Pareto optimization is employed for influence maximization (POIM) where the task of finding a set of influential nodes is reformulated as a bi-objective problem. It has been demonstrated is Pareto optimization is quite effective in coping with NP-hard problems. We theoretically compare POIM with a greedy algorithm, which is a widely used algorithm for mining influential seed nodes. We prove that POIM can obtain the best-so-far theoretically guaranteed approximation performance. Empirical studies verify the theoretical results and show the effectiveness of POIM.