Deep Learning for Stock Market Prediction: A Comparative Study of LSTM and Transformer-Based Models

Stock market prediction involves complicated and dynamic tasks because financial data shows stochastic behavior. The research details an evaluation of the predictive abilities between the Long Short-Term Memory (LSTM) and Transformer-based deep learning models in stock market analysis. Stock market data from the past was employed to test both models for their capability to extract pattern information and generate trustworthy predictive estimations. When applied to stock market predictions the Transformer achieved better results than LSTM since it attained 97.9% accuracy while LSTM only achieved a prediction accuracy of 94.2%. The Transformer-based approach demonstrated superior performance, achieving a lower Mean Absolute Error (MAE) of 8.97, highlighting its capability to better model long-range dependencies in financial data. The research results demonstrate that Transformer models exceed LSTM models in their capacity to handle complex market dependencies and market trend patterns. The study adds value to present-day attempts that use state-of-the-art deep learning strategies to enhance algorithmic trading systems as well as financial business decision systems.