Forecasting COVID-19 Pandemic Using Prophet, ARIMA, and Hybrid Stacked LSTM-GRU Models in India.

Area of Science:

• Epidemiology
• Data Science
• Machine Learning

Background:

• The COVID-19 pandemic (caused by SARS-CoV-2) has severely impacted global public health, society, and economies.
• India is among the most affected nations, facing significant challenges in predicting and managing virus transmission.
• Accurate forecasting models are crucial for public health resource allocation and minimizing transmission risks.

Purpose of the Study:

• To develop and evaluate a predictive analytic technique for COVID-19 cases in India.
• To compare the performance of various forecasting models, including Prophet, ARIMA, and stacked LSTM-GRU.
• To identify the most effective model for predicting confirmed and active COVID-19 cases.

Main Methods:

• Utilized Prophet, ARIMA, and a stacked LSTM-GRU model for forecasting.
• Compared predictions against established models like RNN, GRU, LSTM, linear regression, and polynomial regression.
• Evaluated model performance using metrics such as R-squared and Root Mean Squared Error (RMSE).

Main Results:

• The stacked LSTM-GRU model consistently outperformed other models in predicting COVID-19 cases.
• The proposed model demonstrated superior R-squared and RMSE values compared to ARIMA, Prophet, and individual LSTM/GRU models.
• The stacked model's ability to handle large datasets and abstract features contributed to its enhanced predictive accuracy.

Conclusions:

• The stacked LSTM-GRU model offers a robust and accurate approach for COVID-19 case forecasting in India.
• This predictive capability is vital for understanding future transmission patterns and informing public health strategies.
• The study highlights the potential of advanced deep learning models in managing infectious disease outbreaks.