A novel ITO Algorithm for influence maximization in the large-scale social networks

As a key problem in the social network, Influence Maximization(IM) has received extensive study. Since it is a well-known NP-complete problem, it is a great challenge to determine the initial diffusion seed nodes especially when the size of social network increases. In this paper, we firstly introduce a new index (named Node Key Degree, NKD) to denote the significance degree of each node. A node's NKD is determined by two factors: (1) the number of its direct previous nodes, and (2) the number of its successor offsprings within a certain number of levels. Then, we propose a novel efficient ITO Algorithm to solve the IM problem, termed as ITO-IM. There are three properties and two operators in ITO-IM: the formers include particle's radius, particle's activeness and environmental temperature, the later ones are drift operator and fluctuate operator. During the searching process, the particles in ITO can cooperate with each other to effectively balance the contradictions between exploration and exploitation existing in most of meta-heuristic algorithms. In order to understand the strengths and weaknesses of ITO-IM, we have carried out extensive computational studies on the six real world datasets. Experimental results show that our algorithm achieves competitive results in influence spread as compared with other four state-of-the-art algorithms in the large-scale social networks. (C) 2018 Elsevier B.V. All rights reserved.