Effect of maturity stages on the quality indices of wood apple (Feronia limonia) and modeling of its kinetics by applying machine learning approaches

In the present investigation, an inexpensive and non-destructive method was tested for the appropriate maturity classification of wood apple (Feronia limonia). The investigation was conducted to establish the pronounced effect of maturity stages on the growth kinetics, physico-chemical properties, and other quality indices of wood apple. A systematic trend was observed for all the properties namely sphericity, bulk density (g/cm3), true density (g/cm3), pH, total soluble solids TSS (°Brix), titratable acidity (%) and TSS/TA ratio, etc. of the fruit. In contrast, regular changes were also observed in the color properties at various maturity stages of the wood apple. The maturity kinetics was formulated by applying recurrent neural network (RNN) in compliance with K means cluster algorithm. RNN modeling was applied by considering color property (redness value) as input and six maturity indices as the output of the formulated structure. The RNN architecture, 1-6-6 showed the best results for forecasting the wood apple maturity based on color features. Further, based on the results of the K means cluster algorithm, the maturity stages were classified into three main categories, illustrated in the form of a simplified color chart. Hence, this investigation can be useful for proper control and identification of wood apple maturity during the processing.