Welcome to Chinese Agricultural Science Bulletin,

Chinese Agricultural Science Bulletin ›› 2024, Vol. 40 ›› Issue (27): 126-133.doi: 10.11924/j.issn.1000-6850.casb2024-0036

Previous Articles     Next Articles

Prediction Model of Sensory Quality of Cigar Raw Materials Based on BP Neural Network

HOU Bingqing1(), WANG Shuoli2, ZHANG Youjie1, CAO Yang1, SHI Xiangdong2, DING Songshuang2, LIU Bingyang2, WANG Yihui1()   

  1. 1 Shandong China Tobacco Industry Co., Ltd., Jinan 250014
    2 College of Tobacco Science, Henan Agricultural University/ National Tobacco Cultivation and Physiology and Biochemistry Research Center/ Key Laboratory for Tobacco Cultivation of Tobacco Industry, Zhengzhou 450046
  • Received:2024-01-15 Revised:2024-04-02 Online:2024-09-25 Published:2024-09-24

Abstract:

By establishing BP neural network model, the relationship between conventional chemical components and sensory quality of cigar raw materials was explored, in order to predict sensory quality of cigar raw materials quickly and accurately. With the content of conventional chemical components in cigar leaves from Sichuan, Hubei, Yunnan, Hunan and Nicaragua as input variables and the sensory quality indexes of cigar raw materials as output variables, BP neural network models with topological structure of 6-9-1 were constructed respectively to predict the sensory quality evaluation results of cigar raw materials. The results showed that in the samples tested, the contents of total sugar, reducing sugar, nicotine and chlorine in cigar leaves from four major producing areas in China were higher than those of Nicaraguan tobacco leaves. Nicaraguan tobacco leaves scored higher in aroma quality and volume of aroma, Sichuan tobacco leaves scored lower in irritation, Hubei cigar leaves scored higher in aftertaste, Yunnan tobacco leaves scored lower in impurity, and Hunan tobacco leaves scored higher in combustibility and gray. In this study, the statistical characteristics of conventional chemical components and sensory quality indexes of cigar tobacco samples were good, basically following normal distribution. The difference between the predicted value and the actual value of the BP neural network model was small, among which the correlation coefficients of aftertaste, irritation, gray and total score were all above 0.9. Among the errors of the predicted and actual values of the training set, verification set and test set, except the error interval of the total score was large, the error interval of the remaining most indexes was more than 85% within the range of 0-0.5. The prediction model of the sensory quality of cigar materials established by BP neural network has a good fitting effect, which can be used to predict the sensory quality of cigar materials based on conventional chemical components, and promote the innovative development of Chinese cigar.

Key words: cigar raw materials, conventional chemical composition, sensory quality, BP neural network model, prediction model