层次粒化的不确定多态网络重叠社区发现方法研究

项目名称： 层次粒化的不确定多态网络重叠社区发现方法研究

项目编号： No.61503273

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 张泽华

作者单位： 太原理工大学

项目金额： 21万元

中文摘要： 大数据研究的兴起使得现实具有复杂关联的网络社区挖掘问题成为前沿热点。但当前社区挖掘研究多基于经典图论，忽略节点和链接的属性关联。面对不确定多态网络的多源异构、节点隶属模糊、属性信息粗糙等特征，已有方法无法解决这种不确定网络社区发现问题。本项目以社区扩张算法为依托，结合粒计算层次粒化思想，有效利用多态网络的节点属性和链接结构特征，提出大规模多态网络的快速社区启发式算法。根据社区稳定性度量进行社区隶属分析，给出初始社区的粗粒度描述。通过对个体属性及其结构关系分析对网络进行层次粒化，提出基于属性和链接整合的社区稳定性度量来有效评价重叠区域，并构建层次粒化的多态网络重叠社区发现方法来全面准确地描述多态网络的不确定状态，挖掘网络的重叠社区。项目从算法设计、理论证明、性能指标、综合测试多方面进行研究，形成多态网络社区结构粒化的理论和方法体系，顺应网络经济时代需求，推动社区发现研究，拓展粒计算应用领域。

中文关键词： 粒计算；粗糙集理论；层次粒化；不确定多态网络；重叠社区发现

英文摘要： The big data research prompts the researchers to concern about the community mining with complex structure. The study on community mining based on the classical graph theory, ignores the attributes of vertexes and edges and their relationships, and cannot reflect multi-source heterogeneous data, fuzzy structures, rough attributes, uncertain connection as well as the multi dimensional relationships. Specially, it is unable to detect communities with uncertain information. The research based on local community detection algorithm combines with granular computing. With information of nodes and links both with attributes, the project proposes a fast heuristic community seeds algorithm on complex large-scale multimodal network . Through analysis on the individual and their relation structure, hierarchical granulation communities was carried out by community stability measurement, and the initial coarse grained description on the community could be given. With hierarchical analysis on individual attributes and their structural relationships, the community stability measurement based on attributes and links is put forward, which could effectively evaluate overlapping communities. Then the hierarchical granulation method could accurately describe the uncertainty and mine communities on networks. This project researches on the community discovery in aspects of algorithm design, theoretical proof, performance index, comprehensive test. Finally, the granulation theory on multimodal network is build, which could comply with the network economy era, promote community discovery research, and expand the application fields of granular computing.

英文关键词： granular computing;rough set theory;hierarchical granulation;uncertain multimodal network;overlapping community detection