Relatório técnico da modelagem de parâmetros críticos em estações de tratamento de água com inteligência artificial

Abstract

This report details the steps involved in applying regression models using artificial intelligence to estimate the dosages of chemical compounds, primarily chlorine- and fluorine-based derivatives, in a water treatment plant. The criteria for selecting the water treatment plant, as well as the parameters required for model development, are described using algorithms such as Random Forest, XGBoost, and Multilayer Perceptron Artificial Neural Networks in their default configurations. Data partitioning strategies for training, internal validation, and external validation are also presented. Additionally, the report discusses hyperparameter optimization for the predictive models, highlighting improvements in their performance. In all cases, evaluation metrics are provided to assess predictive accuracy, including Mean Absolute Error (MAE), the coefficient of determination (R²), and the Root Mean Squared Error (RMSE). Finally, model validation is performed through an analysis of the predictive performance of the proposed models.