基于挥发性成分定量预测风味茶油掺浸出茶油的偏最小二乘回归模型的建立

Establishment of Partial Least Squares Regression Model for Quantitative Prediction of Flavor Camellia Oil Adulterated with Extracted Camellia Oil Based on Volatile Components

为解决在压榨风味茶油掺浸出茶油的定量预测问题,本文设计高/低两个掺伪梯度,基于挥发性成分,运用Python语言构建定量预测压榨风味(原香和浓香)茶油掺浸出茶油的偏最小二乘回归(PLSR)模型。研究结果表明,高掺伪梯度下压榨原/浓香茶油样本的定量鉴别PLSR模型的平均R2值均达到了0.998,平均RMSE值为1.127/1.166,大部分样本的相对误差集中在0~0.1之间;低掺伪梯度下压榨原/浓香茶油样本的定量鉴别PLSR模型的平均R2值达到了0.956/0.999,平均RMSE值为0.592/0.094,大部分样本的相对误差集中在0~0.15/0~0.02之间。本文所构建的PLSR模型定量鉴别压榨风味茶油掺浸出茶油的准确率较高,压榨浓香茶油掺浸出茶油的定量鉴别效果要好于压榨原香茶油。

To solve the problem of quantitative prediction of pressed flavor Camellia oil adulterated with extracted Camellia oil, based on volatile components, the Partial Least Squares Regression(PLSR) model was constructed to quantitatively predict the pressed flavor(roasted/original) Camellia oil adulterated with extracted Camellia oil using Python language under high/low adulteration gradients. The results revealed that the average R2 and RMSE value of the PLSR model for quantitative prediction of pressed original/roasted Camellia oil adulterated with extracted Camellia oil reached 0.998 and 1.127/1.166 under high adulteration gradients, respectively, the relative error of most samples was concentrated between 0 and 0.1. The average R~2 and RMSE value of the PLSR model for quantitative prediction of pressed original/roasted Camellia oil adulterated with extracted Camellia oil was 0.956/0.999 and 0.592/0.094 under low adulteration gradients, respectively, the relative error of most samples was concentrated between 0~0.15/0~0.02.The PLSR model constructed in the present paper had higher accuracy in quantitative prediction of pressed flavor Camellia oil adulterated with extracted Camellia oil, and the quantitative prediction effect for roasted fragrant Camellia oil was better than that of original fragrant Camellia oil.