Machine Learning Based Prediction and Impact Analysis of Various Lockdown Stages of COVID-19 Outbreak – A Case Study of India

Abstract

Various measures have been taken into account for the virus outbreak. But how much it successes to control outbreak to fights against COVID-19. Machine learning is used as a tool to study these complex impacts on various stages of the epidemic. While India is forced to open up the economy after an extended lockdown, the effect of lockdown, which is critical to decide the future course of action, is yet to be understood. The study suggests Support Vector Machine (SVM) and Polynomial Regression (PR) are better suited compared to Long Short-Term Memory (LSTM) in scenarios consisting of sparse and discrete events. The time-series memory of LSTM is outperformed by the contextual hyperplanes of SVM which classifies the data even more precisely. The study suggests while phase 1 of lockdown was effective, the rest of them were not. Had India continued with lockdown 1, it would have flattened the COVID-19 infection curve by mid of May 2020. With the current rate, India will hit the 8 million mark by 23 October 2020. The SVM model is further integrated with an SIR (Susceptible, Infected and Recovered) model of epidemiology, which suggests that 70% of India’s population is infected by this pandemic during this 8 month and the peak reached in October 2020 if vaccine not found. With increasing recovery rate increases the possibility of decreasing COVID-19 cases. According to the SVM model’s prediction, 90% of cases of COVID-19 will be end in February.