How Are Chemometric Models Validated? A Systematic Review of Linear Regression Models for NIRS Data in Food Analysis

Abstract

Chemometric models play a critical role in the spectroscopic analysis of food, particularly with near-infrared spectroscopy (NIRS), enabling the accurate prediction and monitoring of physicochemical properties. Although chemometric methods have proven to be useful tools in NIRS analysis, their reliability depends on rigorous validation to ensure the rigour of their predictions and their applicability. This systematic review examines validation strategies applied to regression models in NIRS-based food analysis, emphasising the use of cross-validation, external validation and figures of merit (FoM) as key evaluation tools. This comprehensive literature search identified trends in validation methodologies, highlighting frequent reliance on partial least squares (PLS) regression and common flaws in validation methodologies and their reporting. While external validation is considered the best approach, many studies lack it and employ cross-validation methods solely, which may lead to overoptimistic model performance estimates. Furthermore, inconsistencies in the selection and definition of FoM hinder direct comparison across studies. This review underscores the need for increased methodological transparency and rigour in the validation of chemometric models to enhance their reliability.