Transformer-based deep learning models for adsorption capacity prediction of heavy metal ions toward biochar-based adsorbents
Published in Journal of Hazardous materials, 2023
Recommended citation: https://doi.org/10.1016/j.jhazmat.2023.132773
This study compares application two transformer based architures (Tab-Transformer and FT-Transformer) for prediction of removal efficiency from wastewater. The tabular data consisted of 20 input features and 1550 samples. The prediction performance of transformer based architectures was compared with artificial neural networks and tree based algorithms. We found that FT-Transformer excelled in prediction performance because of its efficienct handling of categorical input features which provide contextual information to numerical input features.