Comparative Analysis of XGBoost and Prophet Models for Monthly Dengue Cases Forecasting in Mayangone Township, Yangon, Myanmar: Projections for 2026

Abstract

Dengue fever remains a significant public health challenge in Myanmar, with Yangon Region consistently reporting the highest number of cases. Mayangone Township, an urban area in Yangon, experiences distinct seasonal outbreaks, predominantly during the rainy season (June–August), driven by increased Aedes mosquito breeding following monsoon rainfall. Accurate forecasting is vital for early warning and effective vector control. This study conducted a comparative analysis of XGBoost (gradient boosting machine learning) and Prophet (additive time series model) for predicting monthly dengue cases in Mayangone Township.Yearly national and regional dengue statistics were used to derive monthly incidence estimates for the township from January 2022 to December 2025, integrated with meteorological variables including rainfall, temperature, and humidity. XGBoost utilized engineered features such as cyclical time encodings, lagged cases, and rainy season indices, while Prophet incorporated the same variables with multiplicative seasonality. Models were evaluated using time series cross-validation with R², MAE, RMSE, and MAPE metrics.Both models produced similar annual forecasts for 2026 (XGBoost: 254.62 cases; Prophet: 252.70 cases). Prophet demonstrated superior predictive performance (R² = 0.9869, MAE = 1.92, RMSE = 2.50, MAPE = 9.54%) compared to XGBoost. Rainfall was identified as the most influential predictor in both models. Prophet projected a higher concentration of cases (66.3%) during the rainy season.The findings suggest Prophet is more accurate for forecasting, while XGBoost offers better interpretability of climatic drivers. The 2026 projections indicate continued seasonal risk, emphasizing the need for proactive vector control measures in Mayangone Township.