Numerical analysis and modeling of water quality indicators in the Ribeirão João Leite reservoir (Goiás, Brazil)

<p dir="ltr">This deposit provides the Python notebook and the input dataset used in the study “Numerical analysis and modeling of water quality indicators in the Ribeirão João Leite reservoir (Goiás, Brazil).” The code implements a statistical–computational workflow for parameter selection (VIF, Bartlett and KMO tests, PCA and FA with <i>varimax</i>) and then trains and evaluates machine-learning models to predict three key physico-chemical indicators: turbidity, true color, and total iron. The dataset (SANEAGO) spans January 2018 to June 2023 with samples collected at the TDA1 intake (13 m depth). The pipeline reduces the original set of 19 variables to 7 with cumulative explanatory power up to 86.67%, compares linear regression, decision tree, random forest, XGBoost, and SVM using MAE, MSE, RMSE, and R², and generates all figures and tables produced by the notebook. Materials ensure full reproducibility and can be adapted to other water sources and water-quality time series.</p>