Table 1.
Descriptive statistics of agroecological and management variables used for modeling coffee rust incidence and severity.
Fig 1.
Class distribution before SMOTE application.
Fig 2.
Class distribution after SMOTE application to training data (5,036 instances per class).
Table 2.
Binary logistic regression predicting coffee rust incidence from microclimatic, agronomic, and spatial predictors.
Table 3.
Variance Inflation Factors (VIF) for predictors in the logistic regression model.
Table 4.
Posterior summaries from Bayesian hierarchical logistic regression predicting coffee rust incidence.
Table 5.
Model fit statistics from Leave-One-Out Cross-Validation (LOO-CV) and Widely Applicable Information Criterion (WAIC) for the Bayesian hierarchical logistic regression model.
Table 6.
Performance metrics for supervised learning models predicting coffee rust incidence.
Fig 3.
Receiver Operating Characteristic (ROC) curves for all machine learning models predicting coffee rust incidence.
Higher AUC values indicate stronger discrimination between infected and non-infected plots.
Fig 4.
Precision–Recall (PR) curves for all machine learning models predicting coffee rust incidence.
Higher AUCPR values indicate better tradeoffs between precision and recall in detecting infected plots.
Fig 5.
Calibration curves for supervised learning models.
The diagonal dashed line represents perfect calibration. Brier scores: Logistic Regression (0.182), CatBoost (0.189), Random Forest (0.194), LightGBM (0.195), XGBoost (0.197), SVM (0.199), Naive Bayes (0.208), ANN (0.215).
Table 7.
Computational scalability metrics.
Fig 6.
Partial dependence of coffee rust incidence probability on NDVI.
Fig 7.
Partial dependence of coffee rust incidence probability on daily relative humidity.
Fig 8.
Partial dependence of coffee rust incidence probability on leaf wetness duration.
Table 8.
Comparative impact of environmental predictors on coffee rust incidence probability. Relative increases refer to changes in predicted probability relative to the model baseline, not absolute percentage probabilities.
Fig 9.
SHAP summary plot showing the ranked importance and directional effect of predictors on coffee rust incidence.
Table 9.
Summary of SHAP value interpretations for key predictors of coffee rust incidence.