基于OGD算法的在线移动平均反转策略

Online Moving Average Reversion Strategy Based on OGD Algorithm

为改善非平稳金融市场环境下在线投资组合策略无法实时动态调整的缺点,提出一种OGDMAR策略。基于在线梯度下降(OGD)算法,对在线移动平均反转策略的预测模型进行改进,使预测模型的系数在每次迭代时都可重新调整。在4个经典数据集上进行数值实验,结果表明,与原策略的累积收益相比,改进策略在4个数据集上分别提升了111%、134%、32%和48%,并且在不同的交易成本条件下累积的收益更高。OGDMAR策略具有应对非平稳环境的能力,不仅在累积收益方面有显著提升,而且具有更强的交易成本承受能力。

In order to improve the shortcomings that the online portfolio selection strategy cannot be adjusted dynamically in real time under the non-stable financial market environment,an OGDMAR strategy is proposed.Based on the online gradient descent algorithm,the prediction model of the online moving average reversion strategy is improved,so that the coefficients of the prediction model can be re-adjusted at each iteration.Then,the numerical experiment is carried out on four classic data sets.The experimental results show that the cumulative return of the improved strategy increases by 111%,134%,32%and 48%on the four data sets respectively,compared with the original strategy.In addition,the cumulative return of the improved strategy is also higher,under different transaction cost conditions.In conclusion,The OGDMAR strategy has the ability to cope with non-stationary environments.It not only has a significant improvement in cumulative return,but also has a stronger ability to bear transaction costs.