Abstract: Influence maximization is a significant research direction in social networks. Its main purpose is to get the most influential users to make the range of influence diffusion maximizing. With the coming of big data, the traditional greedy algorithm can not overcome the time problem of influence maximization effectively because of high time complexity for large-scale social networks. This paper proposes the community division to solve influence maximization. Large-scale social networks are divided into smaller community modules using influence probability. Thus, the isolation between sub-communities is also eliminated at greatest degree considering the boundary nodes between communities. To improve the efficiency further, this paper considers an independent influence path as an influence evaluation unit in each community. At the same time, the most influential nodes are found utilizing parallel processing at every community. Finally, the paper verifies the feasibility and efficiency of the proposed algorithm by simulation experiment, which can adapt to the large-scale social networks better.

Key words: social network, influence maximization, community division, influence diffuse

摘要： 影响力最大化问题是社会网络中的重要研究方向，其主要目的是获取社会网络中最有影响力的用户使通过这些用户获得影响传播范围的最大化。随着大数据时代的来临，传统的贪心算法因为复杂度高而不能有效解决大规模社会网络下影响力最大化的时间问题。提出一种基于社区划分的影响力最大化算法，利用影响概率将大规模社会网络分成较小的社区模块，并考虑社区边界节点之间的联系，从而最大程度缩小因社区划分造成的社区间的孤立。为进一步提高算法效率，在每个社区中以影响路径作为影响评估单元，同时对每个社区并行处理以便更高效地获取有影响力的节点。通过仿真实验验证了算法的可行性和高效性，其可以较好地适应大规模社会网络环境。

关键词: 社会网络, 影响力最大化, 社区划分, 影响传播