A comparative study of modified SIR and logistic predictors using local level database of COVID-19 in India

Purpose - Several national- and state-level studies have been predicting the course of the COVID-19 pandemic using supervised machine learning algorithms. However, the comparison of such models has not been discussed before. This is the first-ever research wherein the two leading models, susceptible-infected-recovered (SIR) and logistic are compared. The purpose of this study is to observe their utility, at both the National and Municipal Corporation level in India. Design/methodology/approach - The modified SIR and the logistic were deployed for analysis of the COVID-19 patients' database of India and three Municipal Corporations, namely, Akola, Kalyan-Dombivli and Mira-Bhayander. The data for the study was collected from the official notifications for COVID-19 released by respective government websites. Findings - This study provides evidence to show the superiority of the modified SIR over the logistic model. The models give accurate predictions for a period up to 14 days. The prediction accuracy of the models is limited due to change in government policies. This can be observed by the drastic increase in the COVID-19 cases after Unlock 1.0 in India. The models have proven that they can effectively predict for both National and Municipal Corporation level database. Practical implications-The modified SIR model can be used to signify the practicality and effectiveness of the decisions taken by the authorities to contain the spread of coronavirus. Originality/value - This study modifies the SIR model and also identifies and fulfills the need to find a more accurate prediction algorithm to help curb the pandemic.