The Application of Neural Network Method for the Prediction of the Osmotic Factors of Crookneck Squash

Abstract

ABSTRACT: In this study the reliability of using response surface-neural network method to predict the osmotic dehydration properties of crookneck squash has been investigated. In order to carry out this project, the osmotic solution concentration, the osmotic solution temperature and immersion time were chosen as inputs and solid gain and water loss were selected as outputs of the designed network. The results showed that the optimal points for the artificial neural network parameters such as the number of neurons, momentum coefficient, learning epoch and the rate to predict water loss and solid gain were 15.75, 0.90, 4999.98 and 0.55, respectively. The results also demonstrated that the model was able to forecast water loss and solid gain with R² values equal to 0.967 and 0.890 where relative error values corresponding to each of these factors were estimated at 0.0205 and 0.0872, respectively