An Accurate Model for Covid-19 Positive Cases in India by using Traditional ARIMA and Artificial Neural Networks (LSTM and Bi-LSTM)

Objective: This study focuses on evaluating the accuracy of models that can be identified by taking the data of COVID-19-positive cases in India. Methods: To build the models by using the procedures, Artificial Neural Networks (ANN) and Auto Regressive Integrated Moving Average (ARIMA). The data has been taken for various time periods of (Covid-19 positive cases) from March 2022 to July 2023; March 2022 to Nov. 2022 and from Nov. 2022 to July 2023. The data was collected from the official website of the World Health Organization (WHO). The traditional ARIMA and Long Short-Term Memory (LSTM) deep learning methods were applied to build models for various time periods. Findings: Model performance is being measured with the error parameter (Root Mean Square Error) RMSE (215.74, 100.36 and 127.81) respectively for all the time periods. LSTM is performing better than the ARIMA with a minimum value of RMSE. Novelty: The study has been done for the time periods of Covid-19 positive cases with the help of LSTM, Bi-LSTM and ARIMA methods. The outcome of these methods gave; LSTM is the accurate and best performance model.