Research On Stock Index Futures Trading Strategy Based On Support Vector Machine

Stock index futures can be 'reversely traded',which means stock index futures can be either buying long or selling short,so that it has many functions such as discovering prices and hedging.At the same time,stock index futures trading is highly leveraged and risky,whether the price trend of stock index futures can be predicted accurately with high probability becomes more realistic guiding significance for real operation.There are several traditional prediction methods for the financial data,such as fundamental analysis,technical analysis,time series analysis and so on.However,in the face of nonlinear financial data,these methods have various limitations,and the prediction result is inaccurate.With the development of statistical theory,data mining and artificial intelligence technology,more machine learning methods have been applied to the field of financial data,and achieved good results.Support vector machine,as a research hotspot in the field of machine learning in recent years,is developed from the statistical theory with the least structural risk,and has strong generalization ability.Given the advantages of support vector machine in predicting small samples and nonlinear data,a stock index futures price classification forecasting model is established in this paper.Further,some trading strategies are proposed based on prediction results.This paper takes China's SSE 50 stock index futures as the research object,and divides the future trend of futures prices into two kinds of ups and downs.Then the rise and fall of future trading prices can be predicted based on historical trading data.The specific process is as follows:Firstly,based on the martingale theory,historical data is used to find out the abnormal value of the closing price which is called 'outlier point' data,and filter samples to achieve data cleaning.Secondly,when choosing relevant indicators that affect futures prices,this paper selects 7 basic market indicators and 18 technical indicators.In order to unify the dimensions of the indicators,the data is normalized by the maximum-minimum standardization method.Even more,principal component analysis is used to reduce the dimension of the data to avoid overlapping information between the indicators.Thirdly,in order to improve the accuracy of the classification prediction model,grid search,genetic algorithm and particle swarm optimization algorithm are introduced to optimize the two parameters of support vector machine.After that,the classification prediction model is proposed.Finally,the model obtained from the training set is used to predict the test set samples.The prediction result shows that the classification prediction model based on particle swarm optimization is the best model,and the accuracy of price prediction reaches 62%.Based on the above practices,this paper uses the optimal classification forecasting model to establish a preliminary stock index futures trading strategy.By calculating the various indicator data of the trading strategy through simulation experiments,it is verified that the trading strategy based on support vector machine has a winning rate of 68.29%,and the yield rate outperforms the market in the same period.