An Adaptive Steering-Vector-Based Evolutionary Algorithm for Influence Maximization in Social Networks

Influence Maximization (IM) is to select a subset of nodes from a social network such that the number of nodes influenced by this subset will be maximized. Due to the growing size of social networks, the search space for IM algorithms also expands, the higher computational overhead leads many scholars to explore more effective and efficient IM algorithms. In this paper, we design an adaptive steering vector (ASV) representing the importance of nodes to guide population evolution, and propose a novel meta-heuristic algorithm named ASVEA to solve the IM problem effectively and efficiently. In ASVEA, an efficient node selection method based on ASV is designed and fully harnessed to speed up the population convergence while not losing potentially important nodes. Specifically, based on ASV, we design novel evolutionary operators as well as a local search strategy to search for the high quality seed set. Furthermore, the steering vector updating strategies including local update and global update are designed to enhance the effectiveness of the steering vector. Experimental results concerning influence spread and running time on eight real-world networks demonstrate that the proposed ASVEA strikes a better trade-off between effectiveness (i.e., the number of influenced nodes) and efficiency (i.e., the running time) compared to five representative algorithms.