An ensemble model for forecasting infectious diseases in India

Area of Science:

• Epidemiology
• Data Science
• Biostatistics

Background:

• Time series modeling is crucial for predicting infectious disease outbreaks.
• Accurate forecasting of diseases like dengue and typhoid is essential for public health planning.
• Existing forecasting models may have limitations in accuracy and scope.

Purpose of the Study:

• To develop and evaluate a simple average ensemble model for forecasting infectious disease cases.
• To compare the performance of the ensemble model against individual regression models.
• To analyze the correlation between infectious disease incidence and ecological variables.

Main Methods:

• Utilized monthly data on dengue and typhoid cases (2014-2017) from India's Integrated Diseases Surveillance Programme.
• Applied Support Vector Regression (SVR), Neural Network (NN), and Linear Regression (LR) models.
• Constructed a simple average ensemble model by combining SVR, NN, and LR, optimizing based on Mean Square Error, Root Mean Square Error, and Mean Absolute Error.

Main Results:

• The proposed simple average ensemble model demonstrated superior performance in forecasting accuracy.
• The ensemble method outperformed individual Support Vector Regression, Neural Network, and Linear Regression models.
• Evaluated correlations between infectious disease cases and ecological variables.

Conclusions:

• The developed simple average ensemble model offers improved forecast accuracy for infectious diseases.
• This ensemble approach provides a more reliable tool for predicting dengue and typhoid outbreaks.
• The findings support the use of ensemble methods for enhancing infectious disease surveillance and control.