Neural network approach for modeling the mass transfer of potato slices during osmotic dehydration using genetic algorithm

Abstract

In this study, an approach for designing a neural network based on genetic algorithm has been used to 
model mass transfer during osmotic dehydration of potato slices. The experimental data were obtained 
through a complete randomized design with different osmotic solutions (5, 10 and 15% w/w) and potato 
to solution ratios (1:6, 1:8 and 1:10) at varying temperatures (30, 40 and 60°C) and the best model 
obtained with optimization of a multi-layer perceptron neural network had a mean absolute error of 
0.260, 0.516 and 0.137 for moisture content, water loss and solid gain of osmotically dehydrated slices 
respectively.