Long Short-Term Memory-Based Recurrent Neural Network Model for COVID-19 Prediction in Different States of India

Abstract

In December 2019, the COVID-19 virus was found in Wuhan, China. In early 2020, the virus reached many countries on different continents. The spread of the COVID-19 virus is mainly accelerated by the large population density. As the novel coronavirus has taken the lives of many people in the world, the main focus was to mitigate the effect of this pandemic with precautions. An early prediction of the possible number of infections can help to take sufficient precautions. There are many available tools to track the transmission dynamics of coronavirus. A deep learning algorithm is one of the efficient tools to track transmission dynamics. In deep learning, long short-term memory (LSTM) is an artificial recurrent neural network (RNN), which efficiently deals with time series data without creating gradient problems. In this chapter, an LSTM-based RNN is proposed for the prediction of active cases per day, confirmed cases per day, and cumulative confirmed cases for each stat in India. The dataset for each province in India is extracted from John Hopkin’s, Baltimore, MD, the US publicly available dataset from June 10, 2020, to August 4, 2021. The proposed LSTM model is incorporated with 10 and 100 hidden units for the prediction of active, confirmed cases, and cumulative confirmed cases, respectively. The proposed model has an input step size of two and predicted for 1 day ahead. The provinces in India had undistributed transmission dynamics of COVID-19 due to different geographical areas and population density. The proposed model with same the hyperparameters has efficiently captured the different transmission dynamics in the first and second waves of a pandemic for the different provinces. The experiments carried out in this chapter proved the efficacy and robustness of the proposed LSTM model for the prediction of active cases, confirmed cases, and cumulative confirmed cases for different provinces in India.