Research on Spread Arbitrage Strategy and Optimization Analysis Based on Deep Reinforcement Learning in High-Frequency Trading

This paper explores the implementation of spread arbitrage strategies in high-frequency trading (HFT) environments by constructing an adaptive trading system based on deep reinforcement learning (DRL). The study focuses on four core components: multi-dimensional data integration and preprocessing, reinforcement learning algorithm framework design, trading signal generation and execution mechanism, and strategy optimization with risk control. Acomprehensive high-frequency trading solution is proposed, which incorporates multi-level feature engineering, optimized state space design, refned execution decision-making, and dynamic parameter adjustment. This approach enhances the robustness and profitability of arbitrage strategies, providing both theoretical foundation and practical guidance for the feld of quantitative trading.