Shelf Life Prediction of ‘Royal Gala’ Apples Based on Quality Attributes and Storage Temperature

Phenotypic changes caused by postharvest deterioration of the quality attributes of apples cause substantial economic losses. Thus, strategies for accurate prediction of the shelf life of apples is urgently needed. In each of the three consecutive years from 2016 to 2018, freshly harvested ‘Royal Gala’ apples were stored at 0, 5, 15, and 25°C, respectively. Subsequently, 11 quality attributes were measured at periodic intervals until the end of storage. To screen fewer and more useful indexes, three input datasets were considered: temperature, color value (L*, a*, b*, △E, and C*), weight loss, firmness, titratable acidity, soluble solids content, starch, and reducing ascorbic acid (D1). The key quality attributes were screened by sparse principal component analysis (SPCA) (D2) and correlation analysis (CA) (D3), using shelf life as the output layer of the artificial neural network based on the back propagation (BP ANN) model. The results showed that the correlation coefficients (r) of the predicted and measured shelf life for D1, D2, and D3 were 0.996, 0.997, and 0.993, respectively, while the mean relative errors were 0.071, 0.074, and 0.074, respectively. Meanwhile, the relative percent root mean square (RMS) values were 0.088, 0.092, and 0.112, respectively. The application of SPCA reduced the quality attributes for the input dataset from 12 to 6. Therefore, SPCA-BP ANN was shown to be a useful model for accurate prediction of the postharvest shelf life of ‘Royal Gala’ apples. Additional key words: artificial neural network, correlation analysis, model, sparse principal component analysis, storage