Machine learning for storage duration based on volatile organic compounds emitted from 'Jukhyang' and 'Merry Queen' strawberries during post-harvest storage

초록

Strawberry (Fragaria × ananassa Duch.) is a widely favored horticultural crop renowned for its unique taste and flavor. To develop an accurate predictive model for strawberry freshness, colorimetric data, total soluble solids, titratable acidity, and volatile organic compounds (VOCs) were investigated in fully ripe 'Jukhyang' and 'Merry Queen' strawberries. Data measurements were conducted after 0, 4, 8, and 12 days of cold storage (10 °C, 79% RH). While conventional quality determinants exhibited subtle changes during storage, the extensive data of VOC metabolites proved sufficient for constructing a strawberry freshness predictive model. Of the algorithms evaluated, including orthogonal projections to latent structures (OPLS), random forest, partial least square (PLS) regression, and artificial neural network multi-layer perceptron (MLP), the multi-layer perceptron algorithm yielded the highest accuracy. Utilizing variations in 25 VOCs, a multi-layer perceptron model was developed with an accuracy of test set R2 = 0.999. Compounds such as 1-hexanol, toluene, ethyl isovalerate, ethyl propionate, and isoamyl alcohol were identified as significant biomarkers for MLP freshness prediction model. These findings highlight the potential of VOCs as key indicators for predicting strawberry freshness during storage. © 2024 The Authors

키워드