A machine learning perspective of the impact of COVID-19 on the Indian stock market

Stock markets across the globe were affected by the outburst of COVID-19 in early 2020. This attracted researchers to analyse and understand the implications of such sudden happenings on stock prices, more so by application of latest methodologies that are slowly finding greater relevance in social sciences. This study uses econometric and machine learning techniques to measure the impact of the COVID-19 pandemic and predict the future trend of the stock market in India. This paper attempts to examine the reliability of traditional methods and machine learning techniques to establish their relevance in predicting stock market trends. The study also uses variable perturbation and least absolute shrinkage and selection operator (LASSO) to identify which variables have more significant predictive weightage in the machine learning models. The study reveals that machine learning models outperform econometric models in their predictive power amidst more significant uncertainty. Moreover, a gated recurrent unit (GRU) model is able to capture the stock market dip and gradual recovery much better than a long short-term memory (LSTM) model. The findings of the study reveal that the number of cases and deaths had a significant impact on stock prices and predictive ability to forecast the NIFTY close price.