Integration of chemometrics and metaheuristic algorithms for predicting the total phenol and antioxidant content of red delicious apples via Vis/NIR spectroscopy.
Opis bibliograficzny
Szczegóły publikacji
Streszczenia
Apples are an important dietary source of phenolic compounds that contribute to antioxidant capacity and fruit quality. Conventional chemical assays used to quantify total phenolic (TP) content and antioxidant activity are destructive and time-consuming, limiting their applicability for rapid quality assessment. In this study, visible and near-infrared (Vis/NIR) spectroscopy in the range of 590–950 nm was evaluated as a nondestructive approach for estimating TP and antioxidant capacity in Red Delicious apples. A total of 180 apple samples were analyzed, and reference antioxidant activity was determined using DPPH, ABTS, and FRAP assays. Following outlier analysis, 158 of the 180 samples were retained for model development. Regression models, including multiple linear regression (MLR), partial least squares regression (PLSR), principal component regression (PCR), support vector machine regression (SVMR), artificial neural networks (ANNs), and a deep learning-based artificial neural network (ANNDL), were compared. The ANN and ANNDL models were included primarily for comparative purposes. Among these models, SVMR demonstrated the most robust predictive performance on independent validation data, whereas MLR showed limited predictive capability. To improve model efficiency and reduce spectral dimensionality, SVMR was combined with several metaheuristic wavelength selection algorithms. The results indicate that Vis/NIR spectroscopy, coupled with appropriate regression techniques, has potential for the rapid estimation of antioxidant-related parameters in apples. However, given the limited dataset size and cultivar specificity, the findings should be considered exploratory, and further validation using larger and more diverse datasets is required before practical implementation.
Open Access
Linki zewnętrzne
Identyfikatory
Metryki
Eksport cytowania
Wsparcie dla menedżerów bibliografii:
Ta strona wspiera automatyczny import do Zotero, Mendeley i EndNote. Użytkownicy z zainstalowanym rozszerzeniem przeglądarki mogą zapisać tę publikację jednym kliknięciem - ikona pojawi się automatycznie w pasku narzędzi przeglądarki.
Informacje dodatkowe
| Rekord utworzony: | 28 września 2026 10:23 |
|---|---|
| Ostatnia aktualizacja: | 28 września 2026 10:24 |