Artificial neural network based on microenvironmental parameters for quality prediction of kiwifruit in storage and transportation

Abstract

The microenvironmental factors of kiwifruit are constantly changing over long-distance transportation, but there are no reliable methods for monitoring and predicting quality changes. An artificial neural network (ANN) kiwifruit integrated quality dynamic prediction model based on microenvironmental parameters (i.e., temperature, relative humidity, carbon dioxide, oxygen, and ethylene levels) was developed using the correlation between microenvironmental parameters and changes in quality indicators during storage. The results showed that storage at 4 °C effectively delayed weight loss, increase in cell membrane permeability and soluble solids, as well as decrease in firmness, titratable acid content, ascorbic acid content and color (C* and L* values), delayed the onset of respiratory peaks, lowered ethylene release, and preserved kiwifruit freshness compared to 10 °C and 20 °C. The optimized back-propagation (BP) neural network model had a hidden layer neuron count of 10, and the coefficient of determination (R²) between the predicted and measured values was 0.998, with an average error within ± 5%, which was highly accurate in predicting the overall quality changes of kiwifruit during storage and transportation.