Index Screening and Evaluation Method of Tobacco Quality Based on Multivariate Linear Model
Tobacco quality evaluation is an important reference for tobacco production,procurement of tobacco raw materials and cigarette application.However,there are some problems such as unclear concept and improper use of statistical methods used in the evaluation.In order to establish a scientific and reasonable evaluation method for tobacco quality,this paper took 110 B2F tobacco leaves from 28 tobacco-producing cities in China as an example,analyzed 13 chemical components and e-valuated the sensory quality of smoking leaves,and constructed a linear model for tobacco quality evaluation based on the principle and idea of multiple linear models.The results showed that the total sugar,reducing sugar,sugar-base ratio,starch,nicotine,total nitrogen,leaf potassium,ash and leaf boron of the 13 chemical components had a significant linear relationship with the smoking score.The correlation between chemical components was strong,and there was the problem of multicollinearity.The stepwise linear regression method could effectively modify the multicollinearity and screen out the two e-valuation indexes of total sugar and tobacco leaf boron.The weight of evaluation index could be determined by leading analysis.The linear model could be applied to comprehensive evaluation of tobacco quality in a certain range.
multivariate linear modelindex screeningquality evaluationmulticollinearityprincipal component analysis
万方数据
