Predictive Analysis of COVID-19 Using LSTM and ARIMA Models with Various Orders in India

Abstract

AbstractCOVID-19 pandemic is a difficult challenge to the world. In this situation of crisis, various pioneers are trying different artificial intelligence-based technologies to predict and analyze the spread of the coronavirus disease so that all nations can prevent this pandemic. In this paper, deep learning approaches such as long short-term memory (LSTM), auto-regressive integrated moving average (ARIMA) with various orders have been used to predict the total people who may get affected with COVID-19 in India. The dataset for this analysis has been created by obtaining the data of total confirmed cases, total deaths, active cases, daily deaths and daily cases of India from ‘Worldometer’. The dataset has been created own especially for this research work only after collecting the required data from the above-mentioned website. Root mean squared error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) have been used to calculate the accuracies of the LSTM model as well as ARIMA model. These models have been created by using Python3 programming language and have been evaluated in Jupyter Notebook. Here, ARIMA model with various orders is analyzed to achieve better results in terms of the above-mentioned parameters. In this paper, ARIMA models are compared with respect to its various orders, dependent on the p, d, q values and LSTM with different epochs using which the comparative analysis of accuracy and loss has been calculated depending upon which the best model can be evaluated.KeywordsLSTMRMSEARIMAMAPECOVID-19MAE