项目名称: 基于机器学习的复杂网络社团结构分析及其应用研究
项目编号: No.60873133
项目类型: 面上项目
立项/批准年度: 2009
项目学科: 金属学与金属工艺
项目作者: 卢宏涛
作者单位: 上海交通大学
项目金额: 31万元
中文摘要: 网络社团结构分析是复杂网络研究的一个重要问题,传统的基于模块度最大化的社团结构分析技术主要集中在单部网络的分析上,存在对社团大小有内在要求、不能保证精确求解最优划分和缺少社团划分合理性评价标准等缺点。本项目结合主成分分析、非负矩阵分解、基于样本(Exemplar)的聚类和最近邻聚类等机器学习和模式识别的理论和方法,提出几种基于这些理论和方法的新颖有效的单部网络社团结构分析算法;以图论和矩阵分析相关理论为工具,在单部网络社团结构分析算法的基础上,研究网络的点图线图变换及其对应线图网络的社团结构分析,同时在网络的特征谱框架下研究有向网络,二部网络和超图网络社团结构划分的新算法;利用网络节点可嵌入欧式空间进行特征分析的特性,提出基于社团结构分析技术的新的机器学习算法;研究所提出的算法在现实网络结构分析和模式识别中的应用。
中文关键词: 复杂网络;社团结构;聚类;特征分析;机器学习
英文摘要: Community structure analysis is an important issue in the research of complex networks. The traditional madularity maximization based apporaches mainly focused on the analysis of unipartite networks, which have some drawbacks such as requiring the community size, no ensuring the optimal community partition and short of suitable criteria for community identification. This project aims to develop some novel and efficient algorithms for community structure analysis of unipartite networks based on machine learning and pattern recognition techniques such as principal component analysis, non-negative matrix factorization, examplar-based clustering and nearest neighbor clustering. By making use of graph and matrix theories, this project will ivestigate the line graph conversion of networks and the community structure analysis of line graphs, and investigate the community structure identification of directed networks, bipartite networks and hypergraph networks. New machine learning algorithms based on community structure analysis techniques by embedding network nodes into Euclidean space are to be investigated. The applications of the proposed algorithms to real networks and pattern recognition will also be involved.
英文关键词: Complex networks; Community structure; Clustering; Feature analysis; Machine learning