基于AdaBoost-ElasticNet的套利算法

An arbitrage algorithm based on AdaBoost-ElasticNet

目的:传统套利算法中的配对交易仅限于两只同质股票,并且假设配对的股票股价之间满足特定关系,为打破这种局限性,提出了一种基于AdaBoost-ElasticNet的套利算法。方法:该套利算法首先引入聚类的思想,来打破选取备选股票对时的行业壁垒;然后使用经AdaBoost提升的Elastic-net算法来挖掘聚类所得簇内股票股价中蕴含的套利机会。结果:在同花顺MindGo量化平台上的进行回测,该套利算法在我国近两年A股市场上的收益率是84.48%,最大回撤是2.05%,夏普率是3.13。结论:该套利算法在取得高收益率的同时降低了最大回撤和提高了夏普比率,表明了该套利算法具有较强的实战价值。

Aims:There is a specific prior relationship between two homogeneous stocks in traditional pairs trading strategy which is one form of arbitrage algorithms.We proposed an arbitrage algorithm based on AdaBoost-ElasticNet algorithm to break this limitation.Methods:Firstly,stocks of different industries were selected by clustering to break the industry barrier.Then the Elastic-net algorithm improved by AdaBoost algorithm was used to find the quantitative relation in multi-stocks.Results:According to an evaluation on two years of historical data of Chinese A-share on the RoyalFlush MindGo,the proposed method reached a cumulative return of 80.48%,a max-drawdown of2.03% and a sharper ratio of 3.13.Conclusions:Experimental results show that the proposed arbitrage algorithm has a higher return with a lower max-drawdown ratio and a higher sharper ratio,which indicates a good practical value.