基于多维标度和聚类的CPI数据结构分析  

Structure Analysis of CPI Data Based on Multidimensional Scaling and Clustering

作　　者：马慧 魏立力 MA Hui;WEI Li-li(School of Mathematics and Statistics,Ningxia University,Yinchuan 750021,China)

机构地区：[1]宁夏大学数学统计学院,宁夏银川750021

出　　处：《兰州文理学院学报（自然科学版）》2019年第3期13-17,共5页Journal of Lanzhou University of Arts and Science(Natural Sciences)

基　　金：宁夏研究生教育创新计划示范课程建设项目(YKC201603);2018年度宁夏大学创新创业教育类教学改革项目(CXJG201819201819)

摘　　要：基于3种距离结构应用多维标度法和Ward聚类法对2016年我国各地区的CPI数据进行分析.利用R软件分析得到Stress值、低维空间匹配图、聚类树图及组合MDS和聚类结果图,研究了8个指标之间的潜在关系.结果表明:(1)医疗保健和家庭设备用品及服务分别与其他6个指标具有较弱的相关性,它们各自构成一类;(2)娱乐教育文化可以与交通和通信、烟酒及用品、衣着分为一类,也可以将娱乐教育文化与居住、食品分为一类.Based on three distance structures, the CPI data of various regions in China in 2016 were analyzed by multi-dimensional scaling method and Ward clustering method. Stress value, low-dimensional spatial matching graph, clustering tree graph, combined MDS and clustering result graph were obtained by R software analysis. Potential relationships among the eight indicators were studied. The results showed that:(1) Health care and household equipment supplies and services have weak correlation with the other six indicators, and they each constitute one cluster;(2) Entertainment education culture can be divided into one category of transportation and communication, tobacco and alcohol supplies and clothing, and can also be classified into residential and food categories.

关 键 词：多维标度法 Ward聚类法 CPI 距离矩阵 应力函数 

分 类 号：O212.4[理学—概率论与数理统计]