Forecasting the trend of tuberculosis incidence in Anhui Province based on machine learning optimization algorithm,...

Area of Science:

• Public Health
• Epidemiology
• Data Science

Background:

• Tuberculosis (TB) is a major global communicable disease concern.
• Accurate prediction of TB incidence remains a significant challenge.
• Effective forecasting is vital for resource allocation and intervention strategies.

Purpose of the Study:

• To develop and validate a novel time-series analysis tool for predicting TB incidence trends.
• To improve the accuracy of TB incidence forecasting using multi-source data fusion.
• To establish a new benchmark for infectious disease trend prediction and influencing factor analysis.

Main Methods:

• Implemented Random Forest (RF), Recursive Feature Elimination (RFE), and LASSO for feature selection.
• Developed an integrated model: RF-RFE-LASSO with Particle Swarm Optimization and Multi-Input Long Short-Term Memory (RRL-PSO-MiLSTM).
• Utilized multi-source data for time-series analysis and prediction of TB incidence in Anhui Province (2013-2023).

Main Results:

• The RRL-PSO-MiLSTM model demonstrated superior prediction performance compared to conventional models.
• Achieved excellent results on the test set: MSE: 42.3555, MAE: 59.3333, RMSE: 146.7237, MAPE: 2.1133, R²: 0.8634.
• The model effectively integrated multi-source data for robust time-series forecasting.

Conclusions:

• The RRL-PSO-MiLSTM model offers a significant advancement in infectious disease time-series prediction.
• This approach provides a valuable new benchmark for analyzing infectious disease trends and influencing factors.
• The methodology serves as a reference for public health incidence rate prediction.