Groundwater is an important resource, but contamination-related safety threats and the need for accurate and consistent evaluation are substantial. Therefore, in the current research, groundwater samples ( n = 237) were collected during the pre-monsoon season (May–June 2024) from Raipur District, Chhattisgarh, Central India, and evaluated drinking water quality using eleven physico-chemical parameters through a Water Quality Index (WQI), the Entropy Water Quality Index (EWQI), and a newly developed Self-Organizing Map-based Water Quality Index (SWQI). The adopted quality indexing methods were also validated and compared by replicating the computations using a different dataset sourced from the study area. Analysis demonstrated high reliability of SWQI with precision (0.952), rationality (0.985), robustness (0.977), and versatility (0.918); uncertainty of only 03% from error using Monte Carlo simulations, which is high fidelity and supports the trustworthiness of proposed SWQI method for the assessment of groundwater quality. The present study also used four Machine learning (ML) algorithms including Random Forest (RF), Gaussian Process Regression (GPR), Support Vector Machine (SVM), and Deep Learning (DL) Machine learning (ML) algorithms to predict the groundwater quality indices. Results showed that DL had the best accuracies, in terms of coefficient of determinations (R² = 0.909, 0.960, and 0.985) while predicting WQI, EWQI, and SWQI. The research illustrates that SOM-based indices are valuable for groundwater quality assessment that can be integrated within ML to provide accurate groundwater quality prediction, which can then inform sound water resource management strategies to offer resource sustainability and water safety.
Kumar et al. (Thu,) studied this question.