<p>Adsorption is a promising technique with significant potential for water purification. In this context, the present study examines the adsorption efficiency of poly(6-(ethoxybenzothiazole acrylamide) (PEBTA) in removing high-valent metal ions from aqueous environments, such as Th(IV), As(V), and Hg(II). Structural and chemical characterization of PEBTA is carried out by FT-IR, ¹H-NMR, ¹³C-NMR, TGA, SEM, and EDAX. The PEBTA adsorbent exhibited excellent maximum mono-layer adsorption capacities of 187.6&#xa0;mg/g for Th(IV), 173.1&#xa0;mg/g for As(V), and 159&#xa0;mg/g for Hg(II), as determined through adsorption isotherm studies. The kinetics of the adsorption process is controlled by pseudo-second-order kinetics, and also found to be in accordance with the Langmuir, Freundlich, Temkin, and Sips isotherm models. The thermodynamic parameters (ΔH°, ΔS°, and ΔG°) demonstrate that the adsorption is exothermic and spontaneous. Often, the experimental characterization of such polymeric materials fails to provide a complete understanding of their behavior due to resource limitations associated with large-scale experimentation. To address such challenges, the presented investigation also developed an XGBoost-driven prediction pipeline capable of effectively predicting the adsorption performance of PEBTA. The developed ML model efficiently captures the inherent relationship between the key features, such as temperature, concentration, contact time, pH, and precursor dose and the target response (percentage removal). The prediction mechanism of the developed ML model is carefully assessed by deploying a model-agnostic explanation. Such ML models can play a pivotal role in exploiting the prediction capabilities and simultaneously diminish the requirement of conducting large-scale experiments for complete characterization.</p> Graphical abstract <p></p>