IDENTIFYING INFLUENTIAL SPREADERS IN COMPLEX NETWORKS

Abstract

Influence maximization is the problem of identifying the set of nodes that maximize the size of the outbreak of a spreading process occurring on the network. This problem is important for strategic decisions in marketing and political campaigns. Typically, the problem consists of finding small sets of initial spreaders in large static networks. Due to its computational complexity, the problem can not be solved exactly. Many methods have been proposed to approximate solutions to the influence maximization problem. Here, we first study the effectiveness of proposed methods on a large corpus of real-world networks. We show that simple heuristic methods with low computational complexity can provide comparable solutions to optimization algorithms with high computational burden. Furthermore, we propose a machine learning based approach that combines heuristic methods to increase the performance of provided solutions. Next, we tackle the problem of noise in network structure and dynamics data. We analyze both the individual and combined effects of structural and dynamical noise on the quality of solutions. We show that implementing artificial noise can improve the performance of optimization algorithms to identify influential spreaders. We further analyze the influence maximization problem on temporal networks. We show that losing the information on the ordering or the timing of the interactions significantly decreases the ability to identify influential spreaders. Furthermore, information of the network structure during the first phases of the spreading dynamics is important in order to successfully find influential spreaders, especially when the recovery probability is high.