多维异构网络上的边和社团的预测与演化的研究

发布时间：2018-11-10 18:52

【摘要】：网络预测与演化一直是复杂网络分析领域的研究热点，已经有大量的研究成果出现。其中大部分的研究都是基于同构网络（比如好友网络）。但是，绝大多数真实网络都是多维异构网络，即点之间存在多种类型的边。以文献合作网络为例，网络中的点是作者，点之间的边不仅仅代表两个作者合作发表过论文，还可以有更深的含义，如共同发表的文章所属的研究方向（如数据库、数据挖掘、机器学习等）。多维异构网络比同构网络具有更强的语义表达能力，在多维异构网络上进行网络的预测与演化研究能得到更好的结果。令人遗憾的是，在多维异构网络上进行网络预测与演化的研究并不多。本文研究的是如何在多维异构网络上处理网络的预测与演化问题。根据研究粒度与研究目的的不同，研究的内容分成三部分，分别是多维异构网络上的连接预测、多维异构网络上的社团演化建模以及多维异构网络上的社团演化预测。在多维异构网络上的连接预测的研究中，本文提出了基于有监督学习的MCBLP算法。基于真实数据（DBLP）的实验验证了MCBLP算法的有效性。在多维异构网络上的社团演化建模的研究中，提出了基于社团演化事件的MCE模型。通过基准实验的测试，，MCE模型可以较为准确地刻画社团的演化。最后，在多维异构网络上的社团演化预测的研究中，本文提出了基于有监督学习的MCP算法。基准实验的测试说明MCP算法可以有效地进行社团的预测。
[Abstract]:Network prediction and evolution has always been a hot topic in the field of complex network analysis, and a large number of research results have emerged. Most of the research is based on isomorphic networks (such as friend networks). However, most real networks are multidimensional heterogeneous networks, that is, there are many types of edges between points. For example, in the literature cooperative network, the points in the network are the authors. The edges between the points not only represent the two authors who have published papers together, but also have a deeper meaning, such as the research direction of the published articles (such as database, data mining, etc.) Machine learning, etc. Multi-dimensional heterogeneous networks have stronger semantic expression ability than isomorphic networks, and better results can be obtained by studying the prediction and evolution of multi-dimensional heterogeneous networks. Unfortunately, there are few researches on network prediction and evolution in multidimensional heterogeneous networks. This paper deals with the prediction and evolution of multi-dimensional heterogeneous networks. According to the difference of research granularity and research purpose, the research content is divided into three parts: connection prediction on multidimensional heterogeneous network, community evolution modeling on multidimensional heterogeneous network and community evolution prediction on multidimensional heterogeneous network. In the research of connection prediction in multidimensional heterogeneous networks, a supervised learning based MCBLP algorithm is proposed in this paper. Experiments based on real data (DBLP) verify the validity of MCBLP algorithm. In the research of community evolution modeling on multidimensional heterogeneous networks, a MCE model based on community evolution events is proposed. The MCE model can accurately describe the evolution of the community through the benchmark experiment. Finally, in the research of community evolution prediction on multidimensional heterogeneous networks, a supervised learning based MCP algorithm is proposed in this paper. The benchmark experiment shows that the MCP algorithm can effectively predict the community.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.02;O157.5

【相似文献】