How machine learning boosts the understanding of organic pollutant adsorption on carbonaceous materials: A comprehensive review with statistical insights

The application of machine learning (ML) is promising to solve the difficulty of predicting the adsorption of various organic pollutants on carbonaceous materials. This study highlights how ML advances the adsorption research, emphasizes the robust model construction specialized for presenting various application scenarios of adsorption models. We introduce, for the first time, a systematic data preparation workflow tailored for optimizing adsorption studies. Emphasis is given to addressing key challenges in data preparation, including managing adsorption datasets, preventing data leakage, and choosing descriptors wisely. Various algorithms used in 39 previous related studies were included in statistical analysis, and the applications of emerging algorithms in adsorption were prospected. For the data-driven model, the application of importance analysis is beneficial for comprehending adsorption mechanisms, transforming the black-box models into a glass-box ones. It facilitates the identification of primary features governing the adsorption of distinct emerging contaminants and the optimized design of efficient carbonaceous adsorbents. In addition, this review provides prospects for the advanced ML applications in adsorption research, such as its integration with reinforcement learning policies. We also explore the potential of ML in addressing the complexities associated with multi-component adsorption. In sum, this review offers unprecedented illumination into the opportunities and challenges posed by ML in the realm of aqueous adsorption processes.