Forecasting Dengue Hotspots Associated With Variation in Meteorological Parameters Using Regression and Time Series...

Area of Science:

• Environmental Science
• Epidemiology
• Data Science

Background:

• Dengue is a rapidly spreading vector-borne disease, necessitating accurate forecasting for public health.
• Climate change significantly influences vector-borne disease transmission, making climate monitoring crucial for dengue prediction.
• Maharashtra state, India, faces considerable dengue burden, requiring localized forecasting models.

Purpose of the Study:

• To develop and compare various regression and time-series models for forecasting dengue incidences.
• To identify key climatic factors influencing dengue outbreaks in nine cities of Maharashtra.
• To establish an early warning system for dengue outbreaks in the region.

Main Methods:

• Collected 10 years of monthly dengue incidence data and five climatic factors (temperature, humidity, rainfall, wind speed) for nine Maharashtra cities.
• Employed multiple regression models (Random Forest, Decision Trees, Support Vector Regression, Linear, Elastic Net, Polynomial) and time-series models (Holt's, ARIMA, SARIMA, Facebook Prophet).
• Evaluated model performance using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²).

Main Results:

• Humidity and mean maximum temperature showed strong positive and negative correlations with dengue incidence, respectively.
• Mean minimum temperature and rainfall had moderate positive correlations; mean wind speed showed weak negative correlation.
• Random Forest Regression and Facebook Prophet were the best-fit models for most cities, indicating their efficacy in dengue forecasting.

Conclusions:

• Humidity and maximum temperature are critical climate drivers for dengue outbreaks in Maharashtra.
• The developed models, particularly Random Forest Regression and Facebook Prophet, can accurately forecast dengue incidences.
• Mumbai, Thane, Nashik, and Pune are identified as high-risk areas during August-October, enabling targeted interventions and an effective early warning system.