Application of machine learning models for water quality prediction in a coal-mining region: Insights into fluoride and phenol contamination in Dhanbad District, Eastern India.

Water quality degradation in mining-dominated regions poses a significant challenge due to the combined effects of geogenic processes and anthropogenic activities. However, limited studies have integrated hydrogeochemical analysis with data-driven modelling to quantify contaminant-specific impacts in coal-mining regions, particularly for pollutants such as phenol. This study evaluates drinking-water quality and identifies dominant contamination pathways in the coal-intensive Dhanbad district, eastern India. A total of 262 water samples were analysed for physicochemical parameters and trace metals following QA/QC protocols. Hydrogeochemical characterization using Piper and Gibbs diagrams indicates the dominance of the Ca2+-Mg2+-Cl--SO42- facies, influenced by rock-water interaction with localized evaporative and anthropogenic effects. Entropy Water Quality Index (EWQI) results show that 54.96% of samples are unfit for drinking due to elevated fluoride and phenol concentrations, whereas exclusion of these parameters reduces this proportion to 3.44%, highlighting their disproportionate impact. Their selection is justified by high toxicity, low permissible limits, and contrasting geogenic and anthropogenic origins. Spearman's rank correlation analysis (ρ), significant at p < 0.05, indicates weak associations between fluoride and major ions (|ρ| < 0.2), suggesting geogenic control, whereas phenol shows moderate positive correlations with EC and major ions (ρ ≈ 0.24-0.32), reflecting anthropogenic influence. Three machine-learning models Support Vector Regression (SVR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) were developed to predict EWQI, among which XGBoost achieved the best performance (R2 = 0.97 ± 0.02, RMSE=27.20 ± 13.57, MAE=15.45 ± 4.51). This framework supports rapid water-quality assessment and sustainable water-resource management in mining-affected regions.