This study, presented by Boroujerd et al., investigates the use of machine learning models to predict drinking water quality levels in Tehran, Iran. After collecting approximately 400 drinking water samples from 2021 to 2024 across 15 sites in western Tehran, the researchers analyzed 17 unique physicochemical and microbial parameters and calculated the Water Quality Index (WQI) based on a framework created by the Canadian Council of Ministers of the Environment. Overall, the results of the study demonstrated that the Kolmogorov-Arnold networks (KAN) and multilayer perceptron (MLP) are effective tools to accurately predict the WQI. While WQI values ranged from 75 to 86, categorizing water quality as generally fair to good, the spatial-temporal mapping revealed that 71% of the study area had good water quality in 2021-2022, which declined to 50% in 2022-2023. Among the prediction models, the classical approaches (XGBoost, RF, SVM, MLR) were less accurate in their predictions, with XGBoost performing the best (R² = 0.87, RMSE = 0.331). Neural networks performed better, with the MLP model achieving strong predictive accuracy (R² = 0.901, RMSE = 0.286) and the KAN producing even smoother and more stable predictions (R² = 0.953, RMSE = 0.197). As a whole, the article highlights the potential of advanced machine learning and the KAN model for improved water quality index prediction.