Prediction of Moisture Content in Boesenbergia rotunda (L.) Mansf. Using an Artificial Neural Network Model

Abstract

The objective of this research was to develop a method for predicting the moisture content of Fingerroot (Boesenbergia rotunda) dried using a solar dryer, by employing an artificial neural network (ANN) model, which is known for effectively predicting various quantities under complex conditions. Experiments were conducted under general sky conditions in May 2024 to gather input data for the model. The ANN was trained using data from the first four rounds of experiments, comprising 148 data sets, and was tested using data from the fifth round, consisting of 37 data sets. Each data set included measurements of solar radiation intensity, relative humidity of the air, air temperature, and moisture content of the Fingerroot. The ANN model that most accurately predicted the moisture content had a structure of three hidden layers, each with eight nodes. The predicted moisture content from the ANN model was compared to actual experimental values, showing a close match with a predicted final moisture content of 5.65 % (wb) and an experimental value of 5.97 % (wb). The model's performance was evaluated with a Root Mean Square Error (RMSE) of 5.0388 and a Mean Bias Error (MBE) of 4.2015. These results demonstrate that the ANN model can accurately predict the moisture content of solar-dried Fingerroot, providing a reliable tool for optimizing the drying process.