This research project is dedicated to improving the classification of water potability by combining three machine-learning techniques. The study involves a thorough comparison of various classification methods, incorporating multiple thresholds through a variable selection strategy. The main goal is to simplify the input set, aiming to significantly reduce computational costs and model complexity. This streamlined method not only facilitates a more efficient training process but also shortens the duration of model training. To rigorously assess the model’s performance, a K-fold cross-validation is implemented within this framework. This comprehensive approach contributes to the advancement of methodologies for assessing water quality, potentially enhancing the efficiency and reliability of models used for classifying potable water.