茶油掺伪定性鉴别模型的对比分析

为了解决茶油掺伪其他植物油的掺伪量定量预测问题，本研究基于14个特征性脂肪酸和甘油三酯指标，设置高/低两种不同掺伪梯度，运用Python语言构建并对比分析了偏最小二乘回归(PLSR)模型和多元线性回归(MLR)模型用于掺伪茶油掺伪量的定量预测的效果。研究表明，PLSR模型对掺伪茶油的定量预测效果不理想，高掺伪梯度下PLSR模型的平均RMSE值高达1.99,低掺伪梯度下PLSR模型的平均R²值(0.8888)较低，平均RMSE值(0.906 6)较高。除了对棕榈油掺伪量的定量预测效果较差外，在高/低掺伪梯度下MLR模型定量预测能力较强，平均R²值达到了0.999 873/0.993 572,平均RMSE值为0.146/0.136。结果表明MLR模型可用于不同掺伪质量分数和梯度下茶油掺伪不同食用植物油的掺伪量定量预测问题，效果较好。

Comparative Analysis of Qualitative Identification Models for Camellia Seed Oil Adulteration

In this paper, based on 14 characteristic fatty acid and triglyceride indicators, the partial least squares regression (PLSR) model and multiple linear regression (MLR) model have been constructed and compared to solve the problem of quantitative prediction of adulteration percentage of camellia oil adulterated with other vegetable oils using Python language under high and low adulteration gradients. The results showed that the PLSR model is not effective in quantitative prediction of adulterated camellia oil, the average RMSE value of the PLSR model under high adulteration gradient reached 1.99, the average R2 value and average RMSE value of the PLSR model under low adulteration gradient reached 0.8888 and 0.9066, respectively. Apart from the weak quantitative prediction ability for palm oil adulteration, the MLR model has strong quantitative prediction ability under high/low adulteration gradient, whose average R2 value reached 0.999873 and 0.993572, respectively, and average RMSE value reached 0.146 and 0.136, respectively. The results showed that the MLR model constructed in this paper can be used to quantitative predict the adulteration percentage of camellia oil adulterated with other vegetable oils under various adulteration concentrations and gradients, and the effect is quite good.