Forecasting COVID-19 cases in India Using Machine Learning Models

Abstract

COVID-19 pandemic has affected the economy and changed the human way of life, disrupting everyone's mental, physical, and financial well-being. Many of the fastest-growing economies are strained owing to the severity and communicability of the epidemic. Because of the increasing diversity of cases and the resulting burden on healthcare practitioners and the government, therefore, predicting the number of infected COVID-19 cases which could be useful in planning the required hospital resources in the future. In this paper, we focussed on information-led methods of estimating the numbers of COVID-19 confirmed cases in the country and their implications in the future, using different learning models such as Sigmoid modelling, ARIMA, SEIR model and LSTM, for protective measures, such as social isolation or the lockout of COVID-19. . Use of raw data by separating an event from the previous event in order to set the time series. The computation of number of positive incidents, number of re-referred incidents are reliable within a limited range. A data-driven forecasting method has been used to approximate the total confirmed cases in coming months. These LSTM model gave very promising results than other models. Hence, this work would help the decision makers to understand the upcoming of the pandemic trajectory in the country and take necessary actions for the effect of interventions.