Development of an artificial neural network as a tool for predicting the chemical attributes of fresh peach fruits.

Abstract

This investigation aimed to develop a method to predict the total soluble solids (TSS), titratable acidity, TSS/titratable acidity, vitamin C, anthocyanin, and total carotenoids contents using surface color values (L*, Hue and chroma), single fruit weight, juice volume, and sphericity percent of fresh peach fruit. Multiple regression analysis (MLR) and an artificial neural network (ANN) were employed. An ANN model was developed with six inputs and 15 neurons in the first hidden layer for the prediction of six chemical composition parameters. The results confirmed that the ANN model R2 = 974-0.998 outperformed the MLR models R2 = 0.473-0.840 using testing dataset. Moreover, sensitivity analysis revealed that the juice volume was the most dominating parameter for the prediction of titratable acidity, TSS/titratable acidity and vitamin C with corresponding contribution values of 39.97%, 50.40%, and 33.08%, respectively. In addition, sphericity percent contributed by 23.70% to anthocyanin and by 24.08% to total carotenoids. Furthermore, hue on TSS prediction was the highest compared with the other parameters, with a contribution percentage of 20.86%. Chroma contributed by different values to all variables in the range of 5.29% to 19.39%. Furthermore, fruit weight contributed by different values to all variables in the range of 16.67% to 23.48%. The ANN prediction method denotes a promising methodology to estimate targeted chemical composition levels of fresh peach fruits. The information of peach quality reported in this investigation can be used as a baseline for understanding and further examining peach fruit quality.