模型集群分析结合回归向量法筛选烟碱近红外定量模型潜变量数
Determination of latent variable number of NIR quantitative model for nicotine by model population analysis combined with regression vector method
-
摘要:
为建立稳定的近红外定量模型以预测贵州省烟叶中的烟碱含量（质量分数），采用模型集群分析结合回归向量的方法确定烟碱近红外模型的潜变量数，其中将利用偏最小二乘回归模型的回归向量计算得到的稳定性参数S作为辅助指标。结果表明：使用本方法最终选择的潜变量数是5个，通过独立验证集验证，校正均方根误差（RMSEC）与预测均方根误差（RMSEP）也更接近。通过研究解决了在最小交互验证均方根误差（RMSECV）或最大决定系数（Q In order to establish a stable near infrared quantitative model to predict the nicotine content in Guizhou tobacco leaves, model population analysis combined with regression vector method was adopted to determine the optimal number of latent variables for the NIR model, and the model stability parameter （S）, which was calculated by PLS regression vectors, was used as an additional indicator. The results showed that the number of latent variables obtained by this method was 5. Root mean square error of calibration （RMSEC） was closer to root mean square error of prediction （RMSEP） through the verification by independent validation sets. This method solved the problem that the number of latent variables cannot be determined when it was hard to determine the minimum of root mean square error of cross validation （RMSECV） or maximum of coefficient of determination （Q
下载: