The research on lightweight reinforcement learning algorithm optimization for high-frequency trading data

The paper presents a lightweight reinforcement learning algorithm for high-frequency trading data to improve real-time and precision of trading decisions. By efficiently designing an algorithm framework with data preprocessing, feature selection and model optimization strategies, the method effectively reduces the computational complexity while maintaining high return and robustness. Experimental comparison shows that the algorithm outperforms the conventional Q-learning and deep Q network (DQN) in key indicators such as cumulative return, Sharpe ratio and maximum drawdown, and the computation time is greatly decreased, fully meeting the demand in high-frequency trading scenarios. The study demonstrates the application value in the financial field of lightweight reinforcement learning and brings new inspiration for optimizing the algorithm under complex trading conditions.