FinRL: Adaptive Model Selection for Reinforcement Learning in Stock Trading

This paper advances the FinRL-DeepSeek framework by introducing a market-adaptive modeling approach to improve automated stock trading agents. While FinRL-DeepSeek integrates stock prices and financial news using reinforcement learning (RL) and large language models (LLMs), our approach incorporates a dynamic model selection mechanism that adapts to market conditions, such as bull and bear markets determined using Nasdaq index trends. We also replace the single-model structure of FinRL-DeepSeek with multiple specialized models, each tailored to individual stocks. This adaptation enables more precise decision making, improving responsiveness to market fluctuations and enhancing overall trading performance.