Optimization of Parameter Selection for Partial Least Squares Model Development

Abstract

In multivariate calibration using a spectral dataset, it is difficult to optimize nonsystematic parameters in a quantitative model, i.e., spectral pretreatment, latent factors and variable selection. In this study, we describe a novel and systematic approach that uses a processing trajectory to select three parameters including different spectral pretreatments, variable importance in the projection (VIP) for variable selection and latent factors in the Partial Least-Square (PLS) model. The root mean square errors of calibration (RMSEC), the root mean square errors of prediction (RMSEP), the ratio of standard error of prediction to standard deviation (RPD), and the determination coefficient of calibration (Rcal2) and validation (Rpre2) were simultaneously assessed to optimize the best modeling path. We used three different near-infrared (NIR) datasets, which illustrated that there was more than one modeling path to ensure good modeling. The PLS model optimizes modeling parameters step-by-step, but the robust model described here demonstrates better efficiency than other published papers.

Modelling Diverse Soil Attributes with Visible to Longwave Infrared Spectroscopy Using PLSR Employed by an Automatic Modelling Engine

“…It must be stated that the APA approach was firstly tested by Zhao et al (2015) [38] who proved the importance of considering multiple pathways for modelling spectral data. However, there are a few differences between the two approaches while the main one is the initial preprocessing or pretreatments procedure.…”

“…However, there are a few differences between the two approaches while the main one is the initial preprocessing or pretreatments procedure. Zhao et al (2015) [38] tested five combinations of pretreatments: while this study analyzed a set of eight potential algorithms and evaluated each in different mutual combinations resulting in up to 120 valid combinations to be applied separately to the spectroscopic data.…”

On the challenges of using field spectroscopy to measure the impact of soil type on leaf traits

“…There is strong autocorrelation in pseudo-absorption values, so PLSR involves dimensionality reduction, producing orthogonal uncorrelated latent vectors containing the maximum explanatory power in relation to the trait data (Wold et al, 2001). The number of latent variables (nL) used in the PLSR analysis was predicted by minimising the prediction residual error sum of squares (PRESS) statistic (Chen et al, 2004;Zhao et al, 2015). We adopted a leave-one-out cross-validation for each PLSR model.…”

Use of Fluorescence and Reflectance Spectra for Predicting Okra (<i>Abelmoschus esculentus</i>) Yield and Macronutrient Contents of Leaves

“…These procedures are largely known as chemometrics in which efforts are made for developing mathematical and statistical methods to extract relevant, useful and efficient information from raw spectral data. In the optimization of chemometric process, steps such as spectral pretreatment, variable selection and latent factors are highly cases sensitive [24]. Especially in quantitative spectroscopy, single wavelengths selection has been shown to improve precision and accuracy in the calibration process [25].…”