Moisture Prediction in Bird’s Nest Drying with Machine Learning Models

The moisture content plays a pivotal role in determining the quality of dried food products. With the aim of refining moisture estimation accuracy during the drying process and streamlining workflow efficiency and automation, this study investigates machine learning models for predicting moisture levels in bird’s nest products. The proposed model comprises two base models and a multi-layer perceptron (MLP) serving as a meta-model. The MLP architecture encompasses an input layer, two hidden layers employing the rectified linear unit (ReLU) activation function, and an output layer. Genetic algorithm and grid search techniques are utilized to optimize the number of neurons in the hidden layers, ensuring the selection of an effective configuration for the meta-model. Through experimental evaluation, the stacking model demonstrates superior performance compared to other models when applied to the bird’s nest drying dataset, achieving notable metrics such as an \({R}^{2}\) $R^2$ value of 0.972 and an MAPE of 6.421%. Hence, this stacking model combined with MLP exhibits the capability to accurately forecast the moisture content of bird’s nest products during the drying process.