Predictive Modeling of Dengue Outbreaks: A Comparison of Statistical and Machine Learning Approaches

Dengue is a major public health concern in tropical regions like Brazil, where accurate forecasting of outbreaks is essential for effective prevention and control. This study compares four data-driven models—SARIMA, Random Forest, XGBoost, and SVR—to predict dengue incidence in São Carlos, Brazil, using epidemiological and climatic data from 2012 to 2022. The results indicate that the SARIMA model best captured the seasonality and temporal dependencies of dengue cases, while XGBoost stood out among the machine learning approaches by effectively identifying general patterns, including epidemic peaks. The findings highlight the effectiveness of statistical models in capturing seasonal patterns, while machine learning models offer complementary strengths in identifying complex relationships and predicting epidemic peaks.