Large Language Model for Dynamic Strategy Interchange in Financial Markets

With the widespread application of large language model technologies, various industries have adopted this technology to optimize existing solutions, including the field of quantitative trading. Large language models, relying on their rich pretrained knowledge, have achieved significant results in key tasks of quantitative trading such as news analysis, market sentiment judgment, and financial data forecasting. These quantitative strategies based on specific tasks can achieve considerable profits in the financial market. However, the instability of financial markets means it is difficult to maintain a market advantage by relying on a single strategy over the long term. In this paper, we draw on the principles of reinforcement learning to propose a dynamic strategy switching method based on large language models, to adapt to market changes. This method dynamically selects the most suitable strategy for execution based on the current financial market data status and predefined multiple quantitative strategies. Evaluation on two real datasets shows that our method outperforms single-strategy models in several aspects. Furthermore, the application of multi-strategy dynamic switching, combining large language models and reinforcement learning, not only demonstrates superiority but also reveals its great development potential.