In this research, a novel hybrid method MCXGBoost-Bagging-RegPSO is proposed for direction forecasting of the high-frequency Apple Futures’ price and simulation trading. First, a multi-classification method based on the eXtreme Gradient Boosting (XGBoost) is established for Apple Futures price direction classification, while the Regrouping Particle Swarm Optimization (RegPSO) is adopted to optimize the parameters of the movement magnitude levels, XGBoost, and the pre-designed trading rules. Next, a Bagging method is incorporated into the proposed approach to solve the overfitting problem. Then, the proposed method predicts the price movement direction and magnitude level, and a one-year high-frequency trading simulation is executed based on the price direction forecasting results. Finally, several evaluation indicators are used to assess the direction prediction and profitability performances of the proposed method. Experimental results demonstrate that the proposed approach successfully achieved outstanding performance in terms of hit ratio, accumulated return, maximum drawdown, and return–risk ratio. As far as it is concerned, the proposed method could be considered as a useful reference for both intraday investors engaged in high-frequency trading and regulators of the Apple Futures market.

在本研究中，提出了一种新的混合方法 MCXGBoost-Bagging-RegPSO 用于高频苹果期货价格的方向预测和模拟交易。首先，针对苹果期货价格方向分类建立了基于极限梯度提升（XGBoost）的多分类方法，同时采用重组粒子群优化（RegPSO）优化运动幅度水平、XGBoost、预先设计的交易规则。接下来，将 Bagging 方法结合到所提出的方法中以解决过拟合问题。然后，该方法预测价格走势方向和幅度水平，并根据价格方向预测结果执行为期一年的高频交易模拟。最后，几个评价指标被用来评估所提出方法的方向预测和盈利能力。实验结果表明，所提出的方法在命中率、累积回报、最大回撤和回报风险比方面取得了出色的表现。就其而言，所提出的方法对于从事高频交易的日内投资者和苹果期货市场的监管机构都可以视为有益的参考。