A genetic algorithm concerning diversity information for solving the robust influence maximization problem on networks

Abstract

The influence maximization problem is one of the hot research topics in the field of complex networks in recent years. The so-called influence maximization problem is how to select the seed set that propagates the largest amount of information on a given network. In practical applications, networks are often exposed to complicated environments, and both link-specific and node-specific attacks can have a significant impact on the network’s propagation performance. Several pilot studies have revealed the crux of the robust influence maximization problem, but the current work available is not comprehensive. On the one hand, existing studies only consider the case that the network structure is stable or under link-specific attacks, and few researches have concentrated on the case when the network structure is under node-specific attacks. On the other hand, the current algorithm fails to combine the information of the search process well to solve the robust influence maximization problem. Aiming at these deficiencies, in this paper, a metric for evaluating the robust influence performance of seeds under node-specific attacks is developed. Guided by this, a genetic algorithm (GA) maintaining the principle of diversity concern (DC) to solve the Robust Influence Maximization (RIM) problem is designed, called DC-GA-RIM. DC-GA-RIM contains several problem-orientated operators and fully considers diverse information in the optimization process, which significantly improves the search ability of the algorithm. The effectiveness of DC-GA-RIM in solving the RIM problem is demonstrated on a variety of networks. The superiority of this algorithm over other approaches is shown.