Leveraging Long Short-Term Memory (LSTM) for Stock Prediction, Anomaly Detection and Risk Analysis in the Indian Banking Sector

Abstract

The right tracking of the stock market and their anomalies can be recognized. These factors are essential in making the right decisions on the investment as far as the Indian banking industry is concerned. The model uses Long Short-Term Memory (LSTM) networks to forecast the price dynamics of the three biggest banking institutions namely: SBI Bank, HDFC Bank, and ICICI Bank with past OHLCV data covering the years 2021 to 2026. It feeds the data into the LSTM autoencoders to identify anomalies in the data in terms of unusual spikes or anomalies. In essence, it aims to identify any latent patterns of time and in advance demonstrate any red flags. After the model training procedures are done, mean absolute percentage error, annualized return, volatility and Sharpe ratios are then calculated in an attempt to determine the relative performance of the predictive models. It is noteworthy that ICICI Bank has the highest accuracy as a predictive institution with the mean absolute percentage error being 1.86 and SBI has the highest risk-adjusted return as depicted in terms of Sharpe ratio of 0.80. The quantitative results, which are obtained as a result, then allow estimating the reliability and risk tolerance of every institution. The LSTM autoencoder is able to capture salient market events i.e the structural downturn of HDFC in August 2025 and the robust increase in SBI in the beginning of the year 2026. The existence of these anomalies' points to the fact that these models are reading not only into the numerical patterns but also in the exogenous shock in the market hence offering useful information to the investors. Furthermore, this shows that the models take into consideration the real-life market dynamics. In general, the empirical results provide practical investment information. This study helps to provide a sound portfolio optimization model to investors that are interested in making the most out of any portfolio decisions that they had in the Indian banking industry by determining the most predictable banking stocks in the industry, and by delineating areas of and exactly where the nascent risk exist.