同质学术Web空间主题结构研究

Research on the Subject Structure in Homogenous Academic Web Space

本研究在总结现有以共链分析和社会网络分析为主的学术网络局部结构识别方法的基础上，提出了改进的两步式K核分析方法，首次引入了复杂网络中的社区识别算法进行链接网络的分割，并尝试通过适用性评测验证快速聚类算法在同质Web链接网络的主题结构识别方面的有效性。最后的实验结果表明，本研究提出的改进K核分析方法可以有效地发现存在于链接网络中的主题聚类现象；同时研究中引入的快速聚类算法对以93所大学网站进行了聚类并获得六个主题类。通过聚类准确率指标计算，该聚类方法的平均准确率为72％。以上结论证实了本研究中采用的从链接关系度量．数据矩阵构建、到链接网络分析的方法体系是有效的。

In this study, sub-structure discovering methods for academic web based on co-link analysis and social network analysis were summarized, then a new improved two-step k-core method was proposed and an algorithm in complex network for community identification was introduced firstly to link network partition. We try to test the usability of the Fast Algorithm on finding subject-based communities within homogenous academic Web. The experimental result shows that the technique of two-step k-core is capable of discovering subject-communities in academic link network. The 93 university websites were clustered into six classes using the introduced Fast Algorithm and the average accuracy is 72%. Consequently, it can be concluded that a series of strategies applied in this study are workable, including measures on link strength, building of link matrix and methodological frame of link network analysis.