Dengue Early Warning System and Outbreak Prediction Tool in Bangladesh Using Interpretable Tree-Based Machine...

Area of Science:

• Public Health
• Machine Learning
• Epidemiology

Background:

• Dengue fever (DF) is a significant global health threat, particularly in Bangladesh.
• Accurate dengue risk prediction is vital for effective control strategies and early warning systems.
• Identifying key risk factors is crucial for forecasting disease epidemics.

Purpose of the Study:

• To develop an interpretable machine learning (ML) model for dengue early warning and outbreak prediction in Bangladesh.
• To analyze climatic, sociodemographic, and landscape factors influencing dengue transmission.
• To establish a framework for advanced analytical techniques in public health surveillance.

Main Methods:

• Employed high-performance ML algorithms: Random Forests, XGBoost, and LightGBM.
• Utilized comprehensive data from January 2000 to December 2021, including sociodemographic, climate, landscape, and dengue surveillance data.
• Applied hyperparameter optimization and SHapley Additive explanation (SHAP) values for model selection and feature importance analysis.

Main Results:

• Identified nonlinear effects of climatic parameters on dengue risk at specific thresholds.
• Determined optimal climatic conditions for dengue risk: 25-28°C minimum temperature, 32-34°C maximum temperature, 75%-85% humidity, 10mm rainfall, and 12m/s wind speed.
• The LightGBM model accurately forecasted dengue outbreaks, with agricultural land, population density, and minimum temperature being significant drivers.

Conclusions:

• The developed ML model serves as an effective early warning system for dengue outbreaks.
• The study enhances understanding of factors driving dengue epidemics.
• Provides a foundation for sophisticated public health analytical tools and interventions.