Improved Evolution Algorithm that Guides the Direction of Individual Mutation for Influence Maximization in Social Networks

Abstract

The purpose of influence maximization in social networks is to find K nodes as the spread source to activate as many nodes as possible. To improve the efficiency and effectiveness of the classic genetic algorithm in large social networks, the diffusion evaluation function is first proposed to estimate the impact range of seed nodes. Then the individuals are initialized based on the diffusion degree centrality of the node. Adapted a crossover strategy is used to help the evolution algorithm to achieve the purpose of local search. Besides, a direction vector is designed to guide the individual’s mutation. Through experiments on real social networks, the improved evolution algorithm can approximate the state-of-the-art greedy algorithm in the final result while also significantly improving time efficiency.