Stock Prediction using Long Short-Term Memory

Abstract The current study deals with the use of advanced machine learning algorithms in prediction of the stock prices, on the Bank Nifty Index. This index is part of Indian stock exchange which is widely known as National Stock Exchange (NSE) of India. In this study researcher used combination of deep learning techniques with interpretability models. To provide accurate stock price predictions, the Long Short-Term Memory (LSTM) network was implemented in the study. With this the well known interpretability methods, for understanding the predictions of model were used in this study that is SHAP (SHapley Additive exPlanations) values and LIME (Local Interpretable Model-agnostic Explanations. These tools facilitated a deeper understanding of the relationships between financial variables and stock prices. The analysis found that, Equity Share Capital (ESC) and Earnings Per Share (EPS) were the two significant variables affecting stock prices. On the other hand, Return on Assets (ROA) was identified as least impacting variable on stock price movements. The results highlight the significance of explainability in financial forecasting by using an LSTM deep learning model with a combination of interpretability methods such as SHAP and LIME. In this perspective, the research will contribute to a confident and more informed decision-making process in investment. Python programming was used for analysis and model implementations, which further shows how advanced programming tools can be used to deal with challenging financial issues.