Comparative Study of Performances of Banks Using Markov Chain Model and LSTM

Today, the share market has become a significant element of the worldwide economic state and has thus gained the attention of many people. Depending on demand-supply economics, the share prices vary day by day. This paper analyzes and forecasts future stock price changes for four Indian banks namely Axis Bank, ICICI Bank, HDFC Bank, and SBI Bank—by comparing the predictions of Markov chain model and an LSTM model. The data for the study period spans from September 3, 2018, to September 1, 2023. The price change is treated as a random process throughout the study, and it is assumed to be a Markov process. The state space is defined as an increase, stable, or decrease in closing prices. The analysis shows that irrespective of the current closing price of the banks, the distributions of shares "up", "same," and "down" for all the banks favor an upward trend. Also, it was observed that if the closing price was up for the first day for all four banks, then it was expected to return to the same state approximately after two days. Furthermore, three-layer LSTM architecture is applied to capture sequential relationships over time to predict closing prices, using MSE as loss function. Using the Markov chain approach, all banks show an upward trend. Predictions using LSTM model shows slightly an over-fitting situation but the predictions are near to accurate. These predictions were obtained on employing different techniques such as dropouts and optimization.