Guided Genetic Algorithm for the Influence Maximization Problem

Influence maximization is a hard combinatorial optimization problem. It requires the identification of an optimum set of k network vertices that triggers the activation of a maximum total number of remaining network nodes with respect to a chosen propagation model. The problem is appealing because it is provably hard and has a number of practical applications in domains such as data mining and social network analysis. Although there are many exact and heuristic algorithms for influence maximization, it has been tackled by metaheuristic and evolutionary methods as well. This paper presents and evaluates a new evolutionary method for influence maximization that employs a recent genetic algorithm for fixed-length subset selection. The algorithm is extended by the concept of guiding that prevents selection of infeasible vertices, reduces the search space, and effectively improves the evolutionary procedure.