A K-shell Decomposition Based Algorithm for Influence Maximization

• Published in: Engineering the Web in the Big Data Era pp. 269 - 283
• Main Authors: Zhao, Qian, Lu, Hongwei, Gan, Zaobin, Ma, Xiao
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2015
• Series: Lecture Notes in Computer Science
• Subjects: Greedy selection strategy; Influence maximization; Information diffusion; shell Decomposition
• ISBN: 3319198890; 9783319198897
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-19890-3_18

Influence maximization is an issue to find a K-node seed set of influential nodes that can maximize the number of influenced nodes in a social network, where K is a given parameter. A greedy algorithm can approximate the optimal result within a factor of (1-1/e-ε) $$(1-1/e-\varepsilon )$$ , but it is computationally expensive. The degree-based heuristic algorithm is simple, but it is of unstable accuracy without considering propagation characteristics. To address these issues, a k-shell decomposition algorithm(KDA) for influence maximization is proposed under the linear threshold model in this paper. First, we present an improved greedy algorithm(IGA) by discarding some unnecessary calculations. Secondly, the network is decomposed using a k-shell decomposition method to calculate the potential influence of nodes. Finally the nodes with the largest potential influence and the nodes with the largest marginal influence degrees are selected at each step to compose a K-node seed set. The experimental results show that KDA can achieve both high efficiency and high accuracy, compared with the existing representative algorithms.