Predicting incidence of hepatitis E using machine learning in Jiangsu Province, China

Area of Science:

• Epidemiology
• Public Health
• Data Science

Background:

• Hepatitis E poses a growing global health challenge, necessitating accurate incidence prediction for effective healthcare planning.
• Understanding factors influencing Hepatitis E prevalence is vital for disease control strategies.

Purpose of the Study:

• To develop and evaluate a predictive model for Hepatitis E incidence.
• To assess the impact of meteorological factors on Hepatitis E prevalence prediction.

Main Methods:

• Utilized Hepatitis E data from Jiangsu Province (January 2005 - March 2017).
• Compared various machine learning models including ARIMA, GBDT, SVM, LSTM, and Bi-LSTM.
• Incorporated meteorological factors (temperature, water vapor pressure, rainfall) into the Bi-LSTM model.
• Evaluated model performance using Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE).

Main Results:

• The Bi-LSTM model integrating meteorological factors achieved superior performance.
• Achieved state-of-the-art prediction accuracy with RMSE of 0.044, MAPE of 11.88%, and MAE of 0.0377.
• Demonstrated significant improvements in interpretability, learning ability, generalizability, and prediction accuracy.

Conclusions:

• The Bi-LSTM model effectively captures linear and non-linear patterns in Hepatitis E incidence data.
• Meteorological factors significantly enhance the accuracy of Hepatitis E incidence prediction.
• This approach offers a robust tool for public health planning and disease management.