共现矩阵聚类分析研究的技术性问题可归纳为三点：对称矩阵的预处理、聚类分析的距离、SPSS软件的适用性。针对这些问题，提出可采用多维尺度分析( MDS)、社会网络分析( SNA)等方法进行自动分类。基于专利数据，采用三种分析方案( Pearson系数与最长距离聚类结合、Cosine系数与MDS结合、SNA聚类)分别对物流信息化技术的IPC号共现矩阵进行处理。实证结果表明：方案1不适宜进行共现矩阵的聚类分析；方案2和方案3能够更准确地揭示共现矩阵蕴含的技术分类；方案3操作简单、分析结果直观。%Technical issues of cluster analysis in co-occurrence matrix were summarized as three points:pre-processing of the symmetric matrix, the distance in cluster analysis, and the applicability of SPSS. Correspondingly, multidimensional scaling ( MDS) and social net-work analysis ( SNA) were suggested to correct the shortcomings of cluster analysis. Three research solutions ( Pearson coefficient com-bined with furthest distance clustering, Cosine coefficient combined with MDS, SNA cluster analysis) were used to analyze IPC co-occur-rence matrix of logistics information technology respectively. It was shown by the empirical study that the first solution was not appropriate to deal with the co-occurrence matrix, the second and the third solutions revealed the technical cluster behind the co-occurrence matrix more accurately while the third method was easier to operate and understand.
展开▼
