This repository contains codes for the paper entitled " Transformer-based deep learning models for adsorption capacity prediction of heavy metal ions toward biochar-based adsorbents" authored by Zeeshan Haider Jaffari, Ather Abbas, Chang-Min Kim, Jinwoo Kwak, Changgil Son, Yong-Gu Lee, Sangwon Kim, Kangmin Chon, Kyung Hwa Cho, also known as the WEIL group at UNIST, Korea. File name "HMI_data" contains the data set of 1,819 data points used in this study to build this model. Besides, The published article can be found on

Biochar materials have recently received considerable recognition as eco-friendly and cost-effective adsorbents capable of effectively removing hazardous heavy metal ions to aquatic organisms and human health accumulated in aquatic ecosystems. This study accurately predicts the adsorption capacity of biochar materials toward heavy metal ions in aqueous solutions using three deep learning models with a large dataset (1,519 data points). The dataset includes 28 input variables, such as pyrolysis conditions for biochar production (5 features), biochar characteristics (3 features), biochar compositions (9 features), and adsorption experimental conditions (7 features). The applied ML models were evaluated with statistical indicators, and the optimal model was selected. Subsequently, the feature importance of the best-performed model was analyzed by the shapley additive explanations (SHAP). Finally, the two-feature SHAP intraction was utilized as a post-processing technique to investigate the optimized experimental conditions.

If you feel any dificulties in executing these codes, please contact us through email on [email protected] or [email protected]. Thank you