Influenza prediction and holiday effects analysis based on Prophet-LSTM model
10.3969/j.issn.1006-2483.2026.01.002
- VernacularTitle:基于Prophet-LSTM模型的流感节假日效应分析及预测效果研究
- Author:
Wenlin CHENG
1
;
Junjun MAO
2
;
Yizhe WANG
1
;
Jiabing WU
3
Author Information
1. School of Big Data and Statistics, Anhui University, Hefei , Anhui 230601, China
2. Laboratory of Computational Intelligence and Signal Processing, Ministry of Education, Hefei , Anhui 230601, China
3. Anhui Center for Disease Control and Prevention, Hefei , Anhui 230601, China
- Publication Type:Journal Article
- Keywords:
Prophet-LSTM;
Influenza;
Holiday effect;
Prevention and control effect;
Prediction model
- From:
Journal of Public Health and Preventive Medicine
2026;37(1):8-12
- CountryChina
- Language:Chinese
-
Abstract:
Objective To investigate the impact of holiday effects and prevention/control measures on the development characteristics and incidence trends of influenza in Hefei City using a Prophet-LSTM hybrid model, and to validate the applicability of the Prophet-LSTM model in influenza prediction by comparing the performance of different forecasting models. Methods Influenza incidence data from Hefei City (2016–2024) were collected to construct a Prophet-LSTM feature analysis and prediction model to analyze the impact of holiday effects and intervention measures on influenza incidence trends. Comparative models (ARIMA, GRU, and TimeGPT) were established and evaluated on the same test set. Results The data analysis revealed significantly increased influenza incidence during holidays (e.g., New Year's Day, Spring Festival, and National Day), while prevention and control measures led to declining trends. The Prophet-LSTM model demonstrated high consistency between the predicted and actual values, outperforming the comparative models with superior MAE (0.209), MSE (0.195), and IA (0.914), indicating higher prediction accuracy and trend-fitting capability. Conclusion The Prophet-LSTM model effectively captures spatiotemporal characteristics of influenza incidence, exhibits enhanced predictive performance when incorporating holiday effects and intervention measures, and demonstrates significant advantages and application value in influenza forecasting.
