基于主成分分析和聚类分析的烟叶物理特性区域归类-烟草科技.pdfVIP

烟草科技/ Tobacco Science Technology 烟草工艺/ Manufacturing Technology 2014年第 8期（总第325期） 基于主成分分析和聚类分析的烟叶物理特性区域归类 * 郭建华，宋纪真，王广山，奚家勤 ，薛超群，高志强 中国烟草总公司郑州烟草研究院，郑州高新技术产业开发区枫杨街2号 450001 摘要：为了对烟叶物理特性进行区域归类，选取安徽、福建、河南等13个产地的烤烟C3F等级烟叶，分析 烟叶的厚度、叶面密度、填充值等8个物理特性指标，采用主成分分析法提取了5个主成分，并基于综合 得分对 13个产地烟叶进行聚类。结果表明：①提取的5个主成分累计贡献率达到84.02%，厚度、叶面密 度、单叶质量和含梗率4个指标是烟叶物理特性的特征指标；②若以平方欧氏距离5.0作为分类标准，可 以把所有样品分为4类，第一类为黑龙江和山东；第二类为河南和安徽；第三类为福建、湖北和湖南；其他 产地归为一类。 关键词：烟叶；物理特性；主成分分析；聚类分析；区域；归类 中图分类号：TS412 文献标识码：B 文章编号：1002-0861 （2014 ）08-0014-04 Regional Classification According to Physical Properties of Tobacco Leaves Based on Principal Component Analysis and Cluster Analysis * GUO Jianhua, SONG Jizhen, WANG Guangshan, XI Jiaqin , XUE Chaoqun, and GAO Zhiqiang Zhengzhou Tobacco Research Institute of CNTC, Zhengzhou 450001, China Abstract: In order to regionally classify the physical properties of tobacco leaves, B2F grade flue-cured tobacco samples were collected from thirteen producing provinces including Anhui, Fujian, Henan, etc. Eight physical properties (thickness, weight per unit leaf area, filling value, etc.) of the samples were analyzed, five principal components were determined by principal component analysis, and the leaf samples were clustered on the basis of their comprehensive scores. The results showed that: 1）The accumulated contribution rate of the five extracted principal components reached 84.02%, and thickness, weight per unit leaf area, single leaf weight and stem content were the characteristic indexes of physical properties. 2 ）Taking square Euclidean distance 5.0 as the standard, all of the samples were classified into 4 clusters, namely, Cluster Ⅰ, samples from Heilongjiang and Shandong; Cluster Ⅱ, samples from Henan and Anhui; Clust