| Since 1990,China's securities market has experienced a period of develop from nothing and become more and more mature?The traditional subjective investment mode of observing investment through personal intuition and experience is becoming more and more inadequate?Numerous historical experiences have proved that when any subject is studied to a certain extent,it will inevitably produce some quantitative description,as well as investment?With the development of mathematics,computer and machine learning technology,quantitative investment began to enter the field of vision of investors?Quantitative investment is an investment method based on the law of large num-bers?It mainly uses modern statistical methods and the method of quantifying personal investment experience to find all kinds of "probably rate winning" models from the massive historical data,and constructs investment models based on these"probably rate winning" models to conduct a large number of transactions?Quantita-tive investment originated in the United States in the 1970s?At the end of the third quarter of 2019,the management scale of the well-known quantitative investment companies such as Bridgewater,Renaissance and AQR Capital Management have exceeded 60 billion US dollars?Among all kinds of quantitative models,quantitative multi factor hedge model is undoubtedly the most eye-catching star?This kind of model offsets the market risk through stock index futures,options and other hedging means,and only obtains the excess return generated by the model for the market?It not only inherits the advantages of the multi factor model,such as considerable income,wide coverage,large model capacity,but also greatly reduces the withdrawal of the model,with high stability?In recent years,due to the development of machine learning technology and a large number of non-linear relationships in financial data,various machine learning algorithms have been applied to the construction of multi-factor models?The main purpose of this paper is to establish a quantitative multi factor hedge model with stable returns by using genetic programming and xgboost?It is mainly conducted under the guidance of the following two directions:first,it requires that the factors selected in the model should have sufficient explanatory power for the stock market;second,the model established based on these factors should have sufficient accuracy,stability,practicability and strong generalization ability?Based on these two directions,this paper mainly constructs the model from the following aspects:first,mining a considerable number of effective alpha factors?In this paper,through the method of genetic programming algorithm mining and manual mining,a considerable number of effective alpha factors are mined out from 211 million stock data from the beginning of 2007 to the end of 2011,which makes the model of this paper have a strong explanation for the stock market?The other is that xgboost algorithm,a machine learning algorithm that has been developed in recent years,is used in the model construction?This algorithm has the advantages of fast training speed,good regression effect,and not easy to over fit?It can better establish an effective relationship between alpha factor and individual stock return,and improve the accuracy and stability of this model?In this paper,rolling adjustment is also used in the model construction The model of participation is adjusted every 252 trading days,which makes the model adapt to the change of market style faster,and the practicability and generalization ability are greatly enhanced?Thirdly,this paper introduces stock index futures hedging,optimizes the traditional hedging method,makes the model's income more stable and the maximum withdrawal significantly reduced?Based on the above design ideas,this paper finally successfully designs an excel-lent quantitative multi factor hedging model?In the back testing period from January 1,2012 to January 1,2020,the annual return of the model is 15.96%,far higher than the benchmark(CSI 300)annual return of 7.3%,information ratio is 0.348,sharp ratio is 1.060,the maximum withdrawal of the model is 10.49%and the maximum withdrawal of CSI 300 in the same period is 46.7%?Under the extreme market conditions of the stock disaster of 2015 and the unilateral decline in the trade war of 2018,the model can still avoid risks and achieve extraordinary returns(the annual return of 2018 was 20.1%,and the Shanghai stock index fell by 24%over the same period,which was highly practical?... |
PDF Full Text Request