Rough set machine learning informed decision rules for effective adsorption of methylene blue and Congo red dyes by hydrochar

Abstract

The remediation of dye-contaminated wastewater has emerged as a critical challenge in environmental management, driving the need for sophisticated predictive approaches to optimize treatment processes. While machine learning applications in hydrochar-mediated dye removal have proliferated, existing studies have failed to establish universally applicable rules across diverse wastewater matrices. This research addresses this gap through the development of a rough set machine learning (RSML) framework that systematically generates interpretable IF-THEN decision rules for adsorption process optimization. Key attributes identified include solution pH, temperature, and the initial concentration ratio of hydrochar to dye, which are critical for accurate predictions of dye removal efficiency. The model's rule induction capability yielded 4 reducts comprising 52 deterministic rules for Congo red and 9 reducts with 75 rules for methylene blue systems, supplemented by 7 and 18 approximate rules, respectively, to handle boundary conditions. The RSML achieved over 80 % accuracy for both dyes, outperforming the existing 14 classifier models. These findings provide significant implications for establishing scientific rules in future dye removal research using hydrochar adsorption, bridging the gap between theoretical adsorption models and practical water treatment applications.