A methodological exploration of monthly incidence rate of hepatitis B in China: a time series and machine learning analysis

Hepatitis B is not only one of the most severe infectious diseases in the world, but also an important public health issue facing China. This study aims to build an optimal prediction model for the incidence rate of hepatitis B in China and provide accurate prediction for early monitoring and early warning of infectious diseases. We systematically analyzed the incidence trend of hepatitis B in China based on the monthly incidence data from January 2001 to February 2026. We established a series of time series and machine learning models. On this basis, we tried to predict the incidence of hepatitis B in the medium and long term by fitting the optimal discriminant model of LSTM + CNN. The prevalence and incidence rate of hepatitis B in the world and China are decreasing year by year, and China is higher than the global average as well as the mortality rate among elderly people aged 55 and above in China is lower than the global average. The average monthly incidence rate of hepatitis B in China was approximately 0.7 per 10,000 from January 2001 to February 2026. All time series models can capture the trend of incidence rate of hepatitis B. The prediction performance of LSTM + CNN hybrid models (R 2 = 0.603) is superior to traditional time series models (exponential smoothing ARIMA and SVR) and individual machine learning models (CNN, GRU, and LSTM). LSTM + CNN model demonstrates competitive performance to predict the monthly average incidence rate of hepatitis B in China. The hybrid deep learning model (LSTM + CNN) has shown great potential in time series prediction, although further validation is needed. This study explores effective methods for predicting infectious disease time series models, providing a new perspective for infectious disease prediction studies.