Prediction of hepatitis E using machine learning models.

Area of Science:

• Epidemiology and Public Health
• Computational Biology and Bioinformatics
• Machine Learning in Healthcare

Background:

• Infectious disease prediction is crucial for public health interventions.
• Hepatitis E poses a significant public health challenge, necessitating accurate forecasting models.
• Model performance for disease prediction varies across different data series.

Purpose of the Study:

• To compare the predictive performance of three distinct models for Hepatitis E.
• To determine the most appropriate model for forecasting Hepatitis E incidence and case numbers.
• To evaluate Autoregressive Integrated Moving Average (ARIMA), Support Vector Machine (SVM), and Long Short-Term Memory (LSTM) recurrent neural networks.

Main Methods:

• Implemented ARIMA using Python's statsmodels library.
• Utilized MATLAB's libSVM library for Support Vector Machine (SVM) modeling.
• Developed a Long Short-Term Memory (LSTM) recurrent neural network using Keras, incorporating dropout and regularization to mitigate overfitting.
• Trained and validated models using monthly Hepatitis E incidence and case data from Shandong province, China (January 2005 - December 2017).
• Evaluated model performance using Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE).

Main Results:

• Nonlinear models (SVM, LSTM) generally outperformed the linear ARIMA model.
• LSTM achieved the lowest RMSE (0.01 for incidence, 11.75 for cases) and MAPE (15.08% for incidence, 13.6% for cases).
• LSTM also yielded the lowest MAE (0.011 for incidence, 9.984 for cases), indicating superior predictive accuracy.

Conclusions:

• Long Short-Term Memory (LSTM) recurrent neural networks demonstrate the highest suitability for predicting Hepatitis E incidence and case numbers.
• The findings highlight the advantage of nonlinear, deep learning approaches over traditional statistical methods for this specific epidemiological forecasting task.
• LSTM's ability to capture complex temporal dependencies makes it a powerful tool for infectious disease surveillance and prediction.