双层PLS算法及其在近红外光谱分析中的应用

针对近红外光谱数据局部效应显著，变量个数多，且彼此间常存在严重的复共线性，并与样品组分含量呈非线性关系，构建了一种双层非线性偏最小二乘回归 (DNPLSR)算法。它将非线性回归和偏最小二 乘(PLS)相结合，先在外层由PLS从样本数据中提取成分，并实现每对成分间的非线性映射，再在内层实施PLS算法，将外层因变量成分的拟合误差反馈计算转换权向量的增量,进一步修正转换权向量，以使外层所提取的成分对因变量具有更优的解释能力。最后，将该法应用于80个谷物样品的水组分含量与其近红外光谱的定量关系建模，效果良好，显示出很强的学习能力，所建模型的预报性能也优于其他方法。

Double-Layer Partial Least Squares Method and Its Application to NIR Spectroscopic Quantitative Analysis

Aiming at the near infrared spectroscopy(NIR) with local effect sensitivity,numerous predictor variables with serious multicollinearity and having nonlinear quantitative relationship with the chemical compositions from the spectral data,a double-layer partial least squares(DNPLS) algorithm was constructed based on the error feedback-weighting correction.The model based on this proposed algorithm was divided into two parts:the outer part that embedded the nonlinear mapping between each pair of partial least square components into the regression framework of the partial least squares(PLS)method and the inner part that estimated the increment of weight vector by linear PLS method.Subsequently,to increase PLS components interpretative capability,the error-based weights updating procedure in the PLS input outer model was deduced and implemented in the DNPLS regression framework,Finally,the application to the corn sample water content modeling of the proposed DNPLS method was presented with a comparison to some other methods.The DNPLS method not only held a fine learning ability but also improved the prediction performance and steady capability.