NEPSE STOCK INDEX PREDICTION USING LSTM: A Deep Learning Approach Utilizing Recurrent Neural Networks

Abstract

Context: The financial market is characterized by high complexity and volatility. Traditional methods do not adequately capture the nonlinear dynamics, and long-term dependencies present in stock market data, making accurate price prediction difficult. Objective: To develop an innovative stock price prediction platform based on deep learning, using the LSTM algorithm to forecast movements in the NEPSE index and provide a centralized and reliable tool for investors. Method: An LSTM model was implemented using the TensorFlow library for time series analysis. Data processing was performed with pandas. The platform was developed with Flask (backend) and HTML + JavaScript (frontend), allowing for the integration of data processing, prediction generation, and result visualization in a user-friendly web interface. Results: The LSTM model was able to capture long-term dependencies in historical data, generating reliable predictions of market behavior. The interface facilitates the interpretation of results and informed decision-making. Conclusions: The implementation of this AI-based platform improves the methodological approach to stock market prediction in the context of the NEPSE. In addition, it shows potential for optimizing investor decision-making and contributing to the development of predictive models in the Nepalese financial sector.