| How to establishment of the financial market model and obtain excess returns has always been a hot topic in financial research.With the improvement of computer performance in recent ten years,More and more researchers try to use artificial intelligence technology to model,forecast and trade in financial markets.But most studies at the present stage only trade on the basis of predictions.Few studies have proposed a complete quantitative trading system from trading prediction to risk management.Therefore,this paper attempts to propose a novel quantitative trading system including forecasting,capital allocation and risk management.The system uses AdaBoost algorithm to model the financial market and forecast to produce trading decisions.According to the forecasting result of the model,the transaction decision can be made.Then a novel online learning algorithm is used to weigh the transaction decision to obtain the allocation of funds at any time in each asset.Finally,according to the maximum withdrawal of a single asset,the allocation of funds is further processed and the final transaction is judged.The quantitative trading system is composed of three parts: the decision-making layer that generates the transaction decision,the fund allocation layer that gets the allocation of funds and the risk management layer that controls the risk according to the maximum withdrawal.In the Hong Kong stock market,1062 trading days of Hang Seng Index component stocks from April 15,2010 to August 3,2014 were selected to form a sample set.Training optimizes the parameters of quantitative trading system.Then,we extract the daily trading data of five stocks from Hang Seng Index stocks,including China Electric Holdings,Heng An International,Exhibition Real Estate Fund,China Life and Bank of China Hong Kong,from August 4,2014 to October 31,2017,to verify the validity of the quantitative trading system.According to the impact of transaction costs and capital allocation on the results,and taking buy-and-hold as a benchmark,we find the following points:(1)The novel quantitative trading system runs well.(2)When the transaction cost is less than 0.5%,the rate of return of the quantitative trading system is much higher than that of the buy-and-hold strategy.The yield indicators including the cumulative return,Sharp ratio and Stirling ratio are significantly higher than the benchmark over a long period of time,which shows the effectiveness of the strategy in the Hong Kong stock market.(3)The algorithm used by the fund allocation layer can improve the overall investment effect.By comparing the quantitative trading system without capital allocation layer,we find that the existence of capital allocation layer can improve the cumulative return,Sharp ratio and Stirling ratio in all cases.(4)The above findings show that the construction ideas of decision-making level,fund allocation level and risk management level as well as the overall framework of quantitative trading system proposed in this paper are feasible and effective,and can provide new ideas for future investment practice or academic research. |
PDF Full Text Request