Abstract In spectroscopy, multivariate calibrations more than often include a pre-processing step to reduce the effect of unwanted (not Y-related) sources of variability. Because there are many types of background noise, there are many pre-treatment methods. It is therefore tedious to select and/or combine the best pre-treatments. This article proposes to combine several pre-treatments through the use of sequential and orthogonalized partial least squares (SO-PLS), thus leading to a boosting method. The performances and properties of this new method, called Sequential Preprocessing through ORThogonalization (SPORT), are compared to those of a previously published stacking method. SPORT demonstrates very good calibration performances, but also the ability to make significant pretreatment selections.

中文翻译：

---

通过正交化 (SPORT) 进行的顺序预处理及其在近红外光谱中的应用

摘要 在光谱学中，多变量校准通常包括一个预处理步骤，以减少不需要的（与 Y 无关的）变异源的影响。因为背景噪声的种类很多，所以预处理的方法也很多。因此，选择和/或组合最佳预处理是乏味的。本文建议通过使用顺序和正交偏最小二乘法 (SO-PLS) 来组合几种预处理，从而产生一种增强方法。将这种称为通过正交化的顺序预处理 (SPORT) 的新方法的性能和属性与先前发布的堆叠方法的性能和属性进行了比较。SPORT 展示了非常好的校准性能，而且还能够进行重要的预处理选择。