LSTM-Based Deep Learning Approach to Analyze the Impact of COVID-19 on Auto and Pharma Sector Stocks of Nifty 50

The stock market's complexities make it complicated, volatile and nonlinear system, making it difficult to forecast stock values effectively. Stock price swings are heavily influenced by a variety of factors including political climates, economic data, sentiment assessments and worldwide events such as the Covid19 pandemic and the Russo-Ukraine War. Deep learning (DL) approaches have been widely used to analyse stock patterns and predict prices using time series data and sentiment analysis. Notably, “Long Short-Term Memory (LSTM)”, a form of “Recurrent Neural Network (RNN) ”, has emerged as an excellent deep learning technique for anticipating future stock closing values. This study was conducted over three time periods: 1st April 2017 to 19th March 2020 (pre-Covid Pandemic), 20th March 2020 to 31st July 2021 (Covid Pandemic - Phases 1 and 2) and 1st August 2021 to 1st June 2024 (post-Covid Pandemic). Data sets of 5 companies of auto sector and 5 companies of pharma sector were taken from Yahoo Finance. LSTM was used as prediction model with Adaptive Moment Estimation (ADAM), Stochastic Gradient Descent (SDG) and RMSprop as optimizer. Experimentation was done using 100 epochs and batch size of 32. The prediction models were evaluated using “Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE)”. Significant results were obtained using ADM optimizer than SGD and RMSprop. Tata Motors from auto sector achieved least RMSE and MAPE in time frame of before Covid Pandemic (12.25,0.064), during Covid Pandemic (6.69, 0.017) and after Covid Pandemic (42.36,0.022). Similarly, in the pharma sector, Sun Pharma achieved least RMSE and MAPE in time frame of before Covid Pandemic (11.81,0.021), during Covid Pandemic (11.24, 0.013) and after Covid Pandemic (25.12,0.013).