李超, 秦云华, 高锐, 吴家灿, 李娥贤, 叶宏, 张承明, 熊文. 基于因子分析的烤烟香型定量判别及与品种对应研究*[J]. 云南大学学报(自然科学版), 2016, 38(5): 805-812. doi: 10.7540/j.ynu.20160134
引用本文: 李超, 秦云华, 高锐, 吴家灿, 李娥贤, 叶宏, 张承明, 熊文. 基于因子分析的烤烟香型定量判别及与品种对应研究*[J]. 云南大学学报(自然科学版), 2016, 38(5): 805-812. doi: 10.7540/j.ynu.20160134
LI Chao, QIN Yun-hua, GAO Rui, WU Jia-can, LI E-xian, YE Hong, ZHANG Cheng-ming, XIONG Wen. MFA-based research for quantitative determination of flue-cured tobacco and correspondence of varieties[J]. Journal of Yunnan University: Natural Sciences Edition, 2016, 38(5): 805-812. DOI: 10.7540/j.ynu.20160134
Citation: LI Chao, QIN Yun-hua, GAO Rui, WU Jia-can, LI E-xian, YE Hong, ZHANG Cheng-ming, XIONG Wen. MFA-based research for quantitative determination of flue-cured tobacco and correspondence of varieties[J]. Journal of Yunnan University: Natural Sciences Edition, 2016, 38(5): 805-812. DOI: 10.7540/j.ynu.20160134

基于因子分析的烤烟香型定量判别及与品种对应研究*

MFA-based research for quantitative determination of flue-cured tobacco and correspondence of varieties

  • 摘要: 为研究烤烟化学组成与其香型和产区间的关系,通过抽样方法收集了2012年国内14省17个品种128个烟叶样品.参照相关行业标准及文献方法测定影响其品质的117种化学指标,对各指标采用MFA降维处理,因子得分构建Bayes香型定量判别模型,并对香型与产区进行对应分析.结果表明:① 原始指标可提出24个公因子,其对原变量的总方差解释率为85.488%;② 定量判别模型能依据巨豆三烯酮、呋喃、碱等类物质的含量对烟叶样品香型进行很好的预测,回判及预测正确率≥ 97.0%;③ 经因子得分加权的香型与品种对应分析结果与主观经验总体一致,各品种的香型特质排序为:清香:红大≥CB-1;中间香:遵烟6号≥贵烟201>毕纳1号>贵烟4号;浓香:中烟100>中烟101≥粤烟97>NC55.

     

    Abstract: To study the relationship among the chemical composition of flue-cured tobacco and its flavor and yield range,128 tobacco samples from domestic 17 varieties in 14 provinces by the sampling method are collected .Reference to the relevant industry standards and literature method ,the quality of 117 kinds of influence chemical indicators were determined,quantitative flavor Bayes discriminant model was constrcucted by MFA dimension reductioneach,scores construct for each-indicator,and correspondence analysis for the flavor and producing areas was done.The results showed that:① 24 common factors can be proposed by original indicators,the total variance of the original variables was explained as 85.488 percent.②quantitative discriminant model can be used to predict the type of flavor of tobacco samples based on the content of megastigmatrienone,furan,alkali and other substances,the correct rate of back judge and predict are ≥97.0%.③ the analysis results of flavor and variety weighted by the factor scores are overall correspondence with the subjective experience,flavor characteristics of all varieties is sorted as:fragrance:Honghuadajinyuan≥CB-1; intermediate incense:Zunyi 6≥Guiyan 201>Bina-1>Guiyan 4; flavor:Zhongyan 100>Zhongyan 101≥Yueyan 97>NC55.

     

/

返回文章
返回