Applications Of ELM To FTSE China A50 Index Futures Trade

FTSE China A50 index futures is the first stock index futures listed outside China.The preparation of the stock index from Chinese A—shares market.Compared with the domestic futures index A50 index futures which has lower transaction costs,higher leverage,higher liquidity and other characteristics which is an important financial instruments for risk management and investment to participants in financial activities.While the A50 stock index futures is influenced by many complex factors,in addition to the time series of nonlinear,dynamic and high noise characteristics,it is hard to predict its trend.Traditional financial time series analyzing tools cannot approach the inherent law of motion,with the development of computer science,artificial intelligence,data analysis,machine learning algorithm is applied to time series in and achieved remarkable results.The main purpose of this paper is to study the application of extreme learning machine in the FTSE Xinhua A50 index futures trading.Because the extreme learning machine have good nonlinear approximation ability,simple structure,fast computing speed.This paper uses extreme learning machine as the main modeling tools.The sample data include 775 trading days of the A50 index futures closing price data during the period from July 26,2013 to September 23,2016,the first 600 data as the training set,the rest 175 data as the prediction set.For comparison,this paper selects four speculative strategies(A1 to A4)and two(B1 and B2)hedging strategies are calculated and analyzed,including A1 and A2 for general according to the technical index trend trading strategy,A3 and A4 according to the results of extreme learning machine trading strategy.B1 is a trading strategy based on the general hedging principle,and B2 according to the model results(see the text for specific instructions).Since support vector machine has a wide range of applications,this paper also uses SVM as a reference model to calculate and compare.The study found that:(1)Extreme learning machine is better than that of Gauss kernel support vector machine(SVM)in the aspect of model structure,learning ability,prediction accuracy.The root mean square error,the mean square error and the mean absolute error of extreme learning machine model are less than the corresponding error of Gauss kernel support vector regression machine.(2)Extreme learning machine not only can accurately predict the future trend of the FTSE A50 index futures daily settlement price but also can predict the specific points.Accurate prediction not only improve the profitability of trading strategies but also reduce the risk of price fluctuations.(3)Empirical results show that whether it is a speculative strategy or hedging strategy the trading strategy based on the results of the model is better than the other trading strategies mentioned in this paper.The former can not only play a role of risk early warning,but also a substantial increase in revenue.The yields of four speculative strategy were 21.84%,24.16%,140.79%,62.80%.A3 strategy gains are A1,A2 6.4,5.8 times.A4 strategy and A2 strategy's earnings were A1 2.9 and 2.6 times,Among them,the yield of A3 strategy is 6.4 and 5.8 times of A1 and A2 strategy respectively.The yield of A4 strategy is 2.9 and 2.6 times of that of A1 and A2,In the hedging strategy,B2 based on the model and it yields 7.1 times as much as B1(according to the general hedge principle).Thus,the trading strategy based on the extreme learning machine model results in significant advantages.