Stock Market Forecasting: From Traditional Predictive Models to Large Language Models

Stock market forecasting is a complex research problem due to the complexity of the factors influencing stock market trends. This survey provides a comprehensive overview of recent advancements in stock market forecasting, focusing on the impact of large language models (LLMs) in financial analytics. The survey explores the strengths and challenges of feature engineering, ensemble methods, hybrid models, text-based prediction and reinforcement learning. It then presents the transformative impact of LLMs, highlighting their capabilities in utilizing transfer learning and few-shot learning to understand complex financial information, enhancing sentiment analysis, improving portfolio management, and stock forecasting accuracy. A key novelty of this survey lies in presenting comprehensive analysis of the strengths and weaknesses of LLMs for different financial tasks in addition to exploring how LLMs can be combined with machine learning and reinforcement learning approaches to overcome their limitations in handling unstructured data, improving model explainability, and enhancing generalizability. Finally, this survey identifies existing research gaps and limitations, proposing future research directions aimed at improving prediction accuracy and utilizing both LLMs and predictive models’ capabilities in stock market forecasting.