LAPSO-IM: A learning-based influence maximization approach for social networks

Abstract

Online social networks play a pivotal role in the propagation of information and influence as in the form of word-of-mouth spreading. Influence maximization (IM) is a fundamental problem to identify a small set of individuals, which have maximal influence spread in the social network. IM problem is unfortunately NP-hard. It has been depicted that hill-climbing greedy approach gives a good approximation guarantee. However, it is inefficient to run a greedy approach on large-scale social networks. In this paper, a local influence evaluation function is presented for optimizing IM problem. The local influence evaluation function provides a reliable expected diffusion value of influence spread under the linear threshold, independent and weighted cascade models. To optimize local influence evaluation function, a learning automata based discrete particle swarm optimization (LAPSO-IM) algorithm is proposed. LAPSO-IM redefines the update rule of particle’s velocity based on learning automata action to overcome the weakness of premature convergence. The experimental results on six real-world social networks show that the proposed algorithm is more effective than base algorithm DPSO with same the level of efficiency and more time-efficient than IMLA with approximate influence spread.