Water Quality Assessment Using the Random Forest Classification Model

This study presents an automated classification model based on the Random Forest model, applied to the assessment of water. A mixed dataset, combining real-world data from field analyses with synthetic data generated according to regulatory thresholds, was used to ensure balanced training of the model. Stratified cross-validation demonstrated satisfactory performance, yielding an average macro F1 score of 0.98 across all classes. Furthermore, evaluation on a test set composed exclusively of real data revealed a strong predictive capability for the majority classes, with an accuracy of 99%. In addition, a dedicated web platform was developed to automate data entry, provide instant classification, and enable dynamic visualization of results. This tool is intended to assist drinking water service managers in monitoring water quality and making rapid decisions in response to detected non-compliances. The results highlight the value of integrating advanced computational approaches into water quality management, while also underscoring the need for continued research on data balancing and validation under real-world conditions.