Application of principal component analysis,cluster analysis and discriminant analysis in the classification of processing properties of Cuibi No.1 flue-cured tobacco
To further classify the processing performance of Cuibi No.1 tobacco,principal component analysis was conducted on 11 physical characteristic indicators such as single leaf weight and leaf density,as well as 5 chemical component indicators such as total plant alkaloids and total sugars,of 400 tobacco samples(produced in Fujian Province from 2020 to 2023).The comprehensive characteristics of tobacco leaves are reflected from 5 aspects:strength performance,chemical performance,size performance,tearing performance,and moisture absorption performance.A mathematical model for the comprehensive score of tobacco processing performance is established,which is Y=0.415y1+0.210y2+0.173y3+0.116y4+0.085y5.The comprehensive score of processing performance of the sample was calculated.Cluster analysis was conducted based on the comprehensive score,and 400 samples were divided into 4 categories.Discriminant analysis was conducted on 294 unclassified tobacco samples based on cluster analysis of classification numbers.The results of principal component analysis showed that for factors with eigenvalues≥1,a total of 5 factors were extracted,and their cumulative variance contribution rate reached 78.643%.The clustering analysis results indicate that the comprehensive score of Class I ranges from-2.62 to-0.84,with average processing performance,mainly consisting of middle to lower tobacco leaves;The comprehensive score of Class II ranges from 1.24 to 3.07,indicating strong processing performance,mainly consisting of upper tobacco leaves;The comprehensive score of Class III was 0.24～1.23,indicating strong processing performance,mainly consisting of middle to upper tobacco leaves;The comprehensive score of Class IV was-0.83～0.23,with moderate processing performance,mainly for central tobacco leaves.The discriminant analysis results showed that the comprehensive score range of processing performance for each category basically overlapped with the clustering analysis results.The initial grouping was discriminated and cross validated,with an initial grouping accuracy rate of 94.3%and a cross validation grouping accuracy rate of 91.3%.This study provided theoretical and data support for the processing and re roasting of Cuibi No.1 tobacco leaves.
Cuibi No.1flue-cured tobaccophysical and chemical characteristicsprocessing performanceprincipal component analysismathematical modelcluster analysisdiscriminant analysis
NETL
NSTL
万方数据
