利用净分析物预处理法简化苹果糖度预测模型

Simplification of prediction model for apple sugar content using net analyte preprocessing

为了简化苹果糖度预测模型,利用净分析物预处理法(NAP)对苹果近红外光谱进行了预处理,并重建了糖度的偏最小二乘(PLS)预测模型.结果表明,随着预处理过程中所用NAP因子的逐个增加,糖度PLS模型的最佳因子数逐渐减少,甚至可减少至1.研究中,当采用5个NAP因子时,PLS糖度模型可以达到最佳性能,此时模型的最佳因子数为7,校正时的相关系数r2和标准偏差SEC分别为0·92773和0·40658,用于预测时的相关系数r2和标准偏差SEP分别为0·90426和0·44221.与NAP法预处理前的PLS模型相比,其精度虽没有大幅提高,但模型显得更加简洁.

To simplify the prediction model of sugar content, net analyte preprocessing (NAP) was used to preprocess the near infrared (NIR) spectra of apples, and the models were rebuilt. Results show that the number of factors used in PLS model will decrease until to 1 finally as the number of NAP factors increase. The best PLS model was obtained when 5 NAP factors were used in pretreatment, which gave the correlation coefficient (r2) of 0.92773, with the standard error of calibration (SEC) of 0.40658 and the standard error of prediction (SEP) of 0.44221. Although this model doesn't improve precision to a great extent, it requires fewer factors and becomes more economical compared with the model before the NAP pretreatment.