基于动态时间规整的配对交易策略研究

【作者】 王晔；

【导师】 李春琦；

【作者基本信息】 上海财经大学 ， 金融， 2021， 硕士

【摘要】 本文尝试着建立以提高选股能力为核心的配对交易策略,研究股票市场数据可以发现,由于市场存在交易摩擦,股票之间的价格存在着一定的领先滞后效应,传统的配对交易选股策略并没有把这种效应考虑在内,本文则尝试着通过动态时间规整算法把领先滞后效应考虑在内,希望以此来选择相关性比较好的股票,并且希望能够在配对交易中获得比较好的回报率。本文在主要完成了以下三部分工作:第一,验证我国的金融市场上是否存在领先滞后效应。本文采用了线性的方法即格兰杰因果来验证上证50指数成分股之间是否存在领先滞后效应,通过对2012-9-10至2020-12-02之间共计2000个交易日的交易数据进行格兰杰因果检验,发现上证50指数成分股之间确实存在着领先滞后效应。第二,利用动态时间规则算法来捕捉并利用领先滞后效应。通过引入改进的动态时间规则算法,即动态互相关系数法(DCCT)来提高配对交易中选择配对股票的准确性。第三,验证动态时间规整算法在我国市场的有效性。为了检验动态时间规则算法是否有效,本文构建了利用两只股票间的协整关系、利用多只股票间的协整关系和基于布林带通道的三种配对交易策略,检验的结果发现在利用多只股票间的协整关系进行配对交易时,多只股票配对交易与两只股票配对交易产生的收益率和夏普比率相差不是很大。在利用两只股票间的协整关系和基于布林带通道技术这两种配对交易策略中,DCCT选股法能过获得比其他选股方法更高的收益率和夏普比率,从而证明了DCCT选股方法在中国市场上也具有较好的选股能力。更多还原

【Abstract】 This article attempts to establish a paired trading strategy centered on improving stock picking ability.Researching stock market data can find that due to the trading friction in the market,there is a certain leading lag effect in the price of stocks.The traditional paired trading stock selection strategy does not Taking this effect into account,this article tries to take the lead-lag effect into account through the dynamic time warping algorithm,hoping to select stocks with better correlation,and hoping to get a better rate of return in pair trading.This article has mainly completed the following three parts of work:First,verify whether there is a leading-lag effect in my country’s financial market.This article adopts the method to verify whether there is a lead-lag effect between the constituent stocks of the Shanghai 50 Index through a linear method,namely Granger causality,by examining a total of 2000 transactions between 2012-9-10 and 2020-12-02 A Granger causality test on daily trading data revealed that there is indeed a leadinglag effect among the constituent stocks of the Shanghai 50 Index.Second,use the dynamic time rule algorithm to capture and take advantage of the lead-lag effect.Through the introduction of an improved dynamic time rule algorithm,namely the dynamic cross-correlation number method(DCCT)to improve the accuracy of the selection of paired stocks in pair trading.Third,verify the effectiveness of the dynamic time warping algorithm in the Chinese market.In order to test whether the dynamic time rule algorithm is effective,this paper constructs three paired trading strategies that use the cointegration relationship between two stocks,the cointegration relationship between multiple stocks,and the Bollinger band channel.When paired transactions are performed on the cointegration relationship between only stocks,there is not much difference between the return rate and the Sharpe ratio generated by the paired transaction of multiple stocks and the paired transaction of two stocks.In the use of the co-integration relationship between the two stocks and the Bollinger Band Channel technology based on these two paired trading strategies,the DCCT stock selection method can achieve higher returns and Sharpe ratios than other stock selection methods,This proves that the DCCT stock selection method also has a better stock selection ability in the Chinese market.更多还原