Evaluation and Prediction of Key Parameters of Water Quality by Machine Learning in the Coastal Area, Southern Iran

Groundwater is the most reliable source of freshwater in arid and semi-arid regions, where surface water resources are scarce and highly variable. Its role in agriculture, drinking water supply, and ecosystem stability highlights the importance of reliable quality assessment. However, traditional monitoring approaches are often costly, labor-intensive, and spatially limited. Machine learning (ML) techniques offer promising alternatives for accurately predicting groundwater quality using a reduced set of hydrochemical parameters. This study evaluated the application of ML algorithms integrated with the Water Quality Index (WQI) framework to predict groundwater quality in the Minab Plain, Hormozgan Province, Iran. Groundwater samples were collected from 15 wells between 2010 and 2023 and analysed for six physicochemical indicators. Correlation analysis was performed to identify the most influential parameters, followed by the application of Decision Tree (DT), k-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forest (RF) models to forecast WQI under variable input scenarios. WQI values ranged from 29.74 to 107.2, with 95.6% of samples classified as “Good” and 4.3% as “Excellent.” pH showed the strongest correlation with WQI (R2 = 0.98), while Na and total dissolved solids (TDS) exhibited the weakest associations. Among the tested models, DT achieved the highest classification accuracy within this dataset; however, the results should be interpreted as case-study-specific rather than universally generalizable. The findings demonstrate that ML-based WQI modeling provides a robust, potentially cost-effective approach for groundwater quality prediction, supporting sustainable water management in arid and semi-arid regions.