Explainable artificial intelligence for predicting dengue outbreaks in Bangladesh using eco-climatic triggers

Area of Science:

• Epidemiology
• Environmental Science
• Artificial Intelligence

Background:

• Dengue poses a significant public health challenge in Bangladesh due to complex transmission dynamics.
• Understanding eco-climatic influences is crucial for effective dengue prevention and control strategies.

Purpose of the Study:

• To apply artificial intelligence (AI) methods to predict dengue outbreaks in Bangladesh.
• To identify key eco-climatic and socio-demographic determinants of dengue transmission.

Main Methods:

• Utilized high-performance machine learning (ML) models (XGBoost, LightGBM) and explainable AI (XAI) techniques (SHAP, LIME).
• Trained models on climatic, socio-demographic, and landscape data from 2000-2023.
• Predicted dengue cases and early warning trends for 2024-2030.

Main Results:

• XGBoost model demonstrated high predictive accuracy (AUC 0.89, Log Loss 0.64) for dengue outbreaks.
• Key predictors identified include population density, precipitation, temperature, and land-use patterns.
• XAI methods highlighted population density, relative humidity, and minimum temperature as significant factors.

Conclusions:

• Explainable AI (XAI) effectively elucidates dengue outbreak complexities.
• An AI-driven framework provides timely alerts for preventing dengue and other infectious disease outbreaks.
• This approach offers a robust tool for public health interventions in Bangladesh.