Tracing Groundwater Contaminants for Sustainable Development Through Machine Learning Algorithm

Groundwater (GW) contamination and its monitoring are continuously posing a significant challenge in cities like Delhi and its surroundings territories. The study analyzed seven hydro‐chemical parameters such as calcium, chloride, magnesium, pH, potassium, total hardness (TH), and electrical conductivity (EC) for 488 GW samples to calculate Weighted Arithmetic Water Quality Index (WA‐WQI), and further prediction of water quality was performed using the machine learning (ML) approaches. This study employed six ML models, artificial neural network (ANN), random forest (RF),

K

‐nearest neighbors (KNN), decision tree (DT), extreme gradient boosting (XGBoost), and support vector machine (SVM) to predict WA‐WQI. The prediction performance was assessed using

R

2

, root mean square error (RMSE), mean absolute error (MAE), centered RMSE (CRMSE), and mean absolute percentage error (MAPE). The obtained results revealed that ANN demonstrated the highest accuracy with lowest errors in MAE of 4.518, MAPE of 3.276, RMSE of 8.294, and CRMSE of 8.236. The ANN model exhibited the highest

R

2

value of 0.99 among all examined models. The performance evaluation order found was ANN > SVM > XGBoost > RF > KNN > DT, and Taylor's diagram was employed to evaluate the efficacy of each model. The validation results show that ANN models with specific samples yielded promising results with

R

2

value of 0.99, MAE of 3.746, MAPE of 2.550, RMSE of 5.520, and CRMSE of 5.373. The outcomes of the study indicate that ANN is the best performing model, providing a cost‐efficient alternative for predicting GW quality, significantly reducing extensive laboratory testing, offering practical support for real‐time monitoring, and helping policy makers to take decisions for mitigation and improving water quality.