Related Experiment Video
Updated: May 6, 2026
08:20
Author Spotlight: AI-Driven Trypanosome Species Detection from Microscopic Images
Published on: October 27, 2023
2.9K
Hybrid ARIMA-LSTM for COVID-19 forecasting: a comparative AI modeling study
Al Mahmud1, Syed Husni Noor Syed Hatim Noor1, Kamarul Imran Musa2
1School of Dental Sciences, Universiti Sains Malaysia, Kubang Kerian, Kelantan, Malaysia.
Peerj. Computer Science
|September 24, 2025
Summary
Hybrid ARIMA-LSTM models significantly improve pandemic forecasting accuracy compared to traditional ARIMA and deep learning LSTM models. This approach offers more reliable predictive analytics for epidemiology.
Area of Science:
- Epidemiology
- Data Science
- Public Health
Background:
- Pandemics pose significant global challenges requiring accurate forecasting.
- Classical statistical models (ARIMA) struggle with nonlinear pandemic data.
- Deep learning models (LSTM) show promise but require extensive resources.
Purpose of the Study:
- To compare the forecasting performance of ARIMA, LSTM, and hybrid ARIMA-LSTM models.
- To evaluate model accuracy using COVID-19 data from Malaysia.
- To determine the most effective modeling approach for pandemic trend prediction.
Main Methods:
- Utilized autoregressive integrated moving average (ARIMA) and long short-term memory (LSTM) models.
- Developed and tested a hybrid ARIMA-LSTM model.
- Evaluated performance using metrics like MSE, MAE, MAPE, RMSE, RRMSE, NRMSE, and R².
Main Results:
- ARIMA models showed poor performance in capturing pandemic trends.
- LSTM models demonstrated improved forecasting accuracy over ARIMA.
- The hybrid ARIMA-LSTM model consistently yielded the lowest error rates.
Conclusions:
- Hybrid ARIMA-LSTM models offer superior pandemic forecasting accuracy.
- Integrating statistical and deep learning methods enhances predictive analytics in epidemiology.
- Hybrid models are recommended for reliable pandemic forecasting and resource allocation.
