Data-driven machine learning improves prediction of sulfonamide antibiotic adsorption by biochar in aqueous phase

Sulfonamide antibiotics (SAs) have attracted much attention due to their environmental risks to aquatic ecosystems. Biochars (BCs), as excellent adsorbent materials, have been used to remove SAs from aqueous phases. To achieve effective evaluation of adsorption, machine learning (ML) strategies are increasingly being developed. However, no applicable data-driven ML models have been studied to predict the adsorption of SAs by BCs in water. Therefore, this study employed an ML approach based on Wasserstein generative adversarial network (WGAN) data augmentation to predict the adsorption of SAs on BCs in the aqueous phase. The results indicated that the WGAN could generate virtual data highly similar to the original adsorption dataset. By expanding the original data using WGAN, the performance of the extreme gradient boosting model in predicting the adsorption amount improved. This study provides new insights into predicting the adsorption behavior of waste-based BCs for SAs in water environments.