基于因子分析的烤烟香型定量判别及其与产区的对应关系

摘要

抽样采集2012年国内14个省29个地区500个烟叶样品,测定了117种化学指标,采用因子分析进行降维处理,基于因子得分构建了Bayes香型定量判别模型,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83.294％.②定量判别模型能依据巨豆三烯酮、呋喃、生物碱等18种物质的含量对烟叶样品香型进行较好的预测,回判及预测正确率≥95.3％.③经因子得分加权的香型与产区对应分析结果与感官评吸结果总体一致,各省份香型特质排序:清香:云＞闽＞蜀;中间香:渝＞黔＞黑＞鄂;浓香:豫＞皖＞粤＞赣.④豫中部、陕东南烟叶属中偏浓香型,黔西南烟叶属中偏清香型.%To study the relationship between chemical composition and its flavor and growing areas,500 tobacco samples were collected from 14 provinces of 29 regions in 2012.By referring to the relevant industry standards and literature,117 chemical indicators were determined.The Bayes discriminant model was established based on factor scores derived from factor analysis to reduce dimension.Correspondence analysis was conducted between flavor type and tobacco growing areas.Results showed that:① 24 common factors were derived from the original indicators,and the total variance of the original variables explained was 83.294％.② Flavor type could be well determined by quantitative discriminant model based on detected content of megastigrnatrienone,furan,alkaloids and other 18 substances,with precision rate of back fitting was ≥ 95.3％.③ After scores weighted,corresponding analytical results between flavor and growing areas were consistent with sensory evaluation results.Yunnan province was the most prominent in light flavor,followed by Fujian province and Sichuan province.Chongqing municipality was the most prominent in intermediate flavor,followed by Guizhou province,Heilongjiang province and Hubei province..④ Tobacco from central part of Henan province,southeastern part of Shaanxi belong to the intermediate strong-flavor type,and tobacco from part of southwestern Guizhou province belong to the intermediate light-flavor type.