Artificial neural networks for predictive modulation of water vapor permeability of clay-based chitosan films

Abstract

The objective of this work is to apply Artificial Neural Networks (ANN) for the predictive modeling of water vapor permeability of chitosan films with clay, this being one of the most important factors in the characterization of edible films and coatings. For this, 30 random data sets were taken from the database obtained by Casariego et al. (2009), for training selection and validation; and 10, equally occasional with the aim of introducing values ​​outside the training of the Network and comparing the expected values ​​against real results obtained in the database used. The design, programming and validation of the network was carried out in the Neural Designer Computer Software (Version 591). It was replaced as a model, Perceptron Multilayer; and a 3:5:1 architecture (Input Layer: Hidden Layer: Output Layer) using Hyperbolic Tangent as activation function and Levenberg-Marguardt as learning algorithm. An EMC of 0.00046 and an R2 of 0.9207 were obtained in the Regression of the real data against the simulated data, this being a satisfactory result that demonstrates the correct training of the Network and the effectiveness of the prediction.