“…When the uncorrelated and non-linear variables of NIR spectra are removed with feature selection, the prediction ability and robustness of the final model is improved ( Yun, 2022 ). It has also been proven to improve the prediction accuracy in fish ( Currò et al, 2021 ; Lv et al, 2017 ; O'Brien et al, 2013 ; Varrà et al, 2021 ), quinoa flour ( Wang et al, 2022 ), asparagus ( Richter et al, 2019 ), garlic ( Han et al, 2024 ), tea ( Jin et al, 2022 ), mushrooms ( Chen et al, 2022 ) and many other food products. It involves the identification and selection of relevant spectral regions or features that carry the most discriminatory information about the geographical origin of the food.…”