Prediction of Moisture Content in Kiwi (Actinidia deliciosa) Dried Using Machine Learning Approaches

Abstract

Predicting product drying kinetics is crucial for achieving optimal drying processes without compromising product quality. This prediction technique necessitates the development of numerical drying models. The aim of this research is to compare prediction models developed using two popular machine learning approaches in recent years: Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Artificial Neural Networks (ANN). In this study, kiwi slices of three different thicknesses were dried using 90 W microwave power. Prediction models were developed using experimental data. The input for the training algorithm included kiwi slice thickness and drying time, while the output was the moisture content of the product. The performance of the models was evaluated by comparing the obtained outputs with experimental data from test sets. These models were assessed using mean absolute percentage error, correlation coefficient, root mean square error and mean bias error metrics. The ANN-based prediction model demonstrated better performance compared to the ANFIS model. The results of these tests indicate that both methods can be used for predicting the moisture content of kiwi slices.