COVID-19 Infection Dynamics for India-Forecasting the Disease using SIR models

ABSTRACT

A retrospective cohort study of novel Coronavirus disease (COVID-19) on India data and predicting the disease outcome (Infected Recovered) using SIR compartment model.An existing literature survey of SIR modeling on coronavirus disease was probed and further improvised the finding and methodologies used in modeling were performed over the existing India data set. Numerous papers were surveyed and the model was trained to understand the optimal value of hyperparameters (β and γ) and the approach to forecast the coronavirus disease.As on 30th June 2020 a total of 215k Active cases and 17k deaths were reported. A sample(N) of 2% of the overall population (133 cr.) was considered and the initial date of infection to model the disease was 1st May 2020. Further comparison was made using 5% and 10% Susceptible population to check the model efficacy. The reported number of Initial Suspected(S0), Infected(I0) and Recovered(R0) cases were 25,965k, 24,755 9,065 respectively. The optimal values of Beta(β) and Gamma(γ) were estimated to be 0.093 and 0.055 respectively. The overall Infected MAPE(Mean percentage absolute error) was 12.23% Recovered MAPE was 6.48% on 2% Susceptible population.The research presented the current trends of Covid-19 disease in India from 1st May'20 to 30th June'20. The trajectory of the disease was also forecasted for next days from 30th June 2020.Rapid growth of the disease has been seen in the month of June with maximum peak reaching on 164th Day (Mid Oct'20) with 25.5 Lakh infected cases considering 2% Susceptible, 64.75 lakh cases considering 5% Susceptible 1.3 cr. cases considering 10% Susceptible. The estimated R0 (β/γ) value is 1.69. However further research and addition of compartments like Exposed, Critical can be made to improve the forecast. © 2020 IEEE.