摘要
针对现有均值反转类策略存在的预测模型参数无法动态更新和未充分考虑动量效应的问题,提出一种策略M-ODMAR。使用简单移动平均模型对股票价格进行预测,并通过在线牛顿步(Online Newton Step,ONS)算法对模型参数进行动态更新;利用在线被动攻击(Passive Aggressive,PA)算法选取投资组合;使用L1中位数来提取价格动量信息并对投资组合进行调整。实验结果显示,在四个数据集上该策略的累积收益高于所对比的其他策略,说明了参数的动态更新和动量效应的加入对于均值反转类策略的累积收益提高具有促进作用。
The existing mean reversion strategies are faced with the problems that the parameters of the prediction model can t be dynamically updated and the momentum effect is not fully considered.In order to solve these problems,the M-ODMAR strategy is proposed.The stock price was predicted using the simple moving average model,and the model's parameters were dynamically updated through online Newton step(ONS)algorithm.The portfolio selection was conducted by passive aggressive(PA)algorithm.The price momentum information that was extracted by L1-median was used to adjust the portfolio.The experimental results show that the cumulative return of the M-ODMAR strategy is higher than the other comparison strategies on the four data sets,indicating that the dynamic update of parameters and the exploration of the momentum effect can improve the cumulative return of mean reversion strategies.
作者
吴金明
刘钊
Wu Jinming;Liu Zhao(School of Mathematics and Statistics,Shanghai University of Engineering and Technology,Shanghai 201620,China;College of Computer Science and Engineering,Cangzhou Normal University,Cangzhou 061001,Hebei,China)
出处
《计算机应用与软件》
北大核心
2023年第5期305-311,共7页
Computer Applications and Software
关键词
在线投资组合选择
在线牛顿步算法
均值反转
动量效应
简单移动平均
Online portfolio selection
Online Newton step algorithm
Mean reversion
Momentum effect
Simple moving average
