Quantitative Trading Strategy Of Stock Index Futures Based On SVM Theory

As an important part of market economy, stock index future and stock are both known as the barometer of modern economy. However, when investors are investing in the stock index future market, they have to face up with greater risk because of the lever inside stock index future. Therefore, the key of succeed in investing in stock index future is whether someone can precisely predict the tendency of the price of stock index future or not.Generally, some traditional stock market analysis methods are based on series of strict premises, which makes these analysis methods don’t work well in analyzing non-linear securities market. Thanks to the rapid development of computer technology, the analysis methods of security market are more abundant than before. Because of its outstanding performance in non-linear approximation, machine learning is widely used in analyzing security market.The research goal of this article is to build a classification and prediction model which is based on the theory of support vector machines, to take a step futher, a trading strategy aimed at Shanghai Stock Exchange (SSE) 50 index future will be built on the classification and prediction model mentioned before. We hope that this trading strategy will provide references for investors when they are speculating in the stock index future.Because support vector machines have better performance in solving small-n problem and non-linearn problem, it’s more suitable to analyze and predict the SSE 50 index future, which is slightly young. In order to select the best kernel function, we build different models based on different kernel function. By carrying out static simulation based on different input vectors, we find that Gaussian RBF kernel function behaves behaves better in classifying and predicting SSE 50 index future’s price.To make a comparison between the performance of basic market indicators and technical indicators in predicting the close price of SSE 50 index future main contract, we build two different dynamic prediction models and corresponding trading strategys. By utilizing historical data to back-test the two strategys mentioned before, it turns out that although the strategy based on basic market indicators behaves far more better than the strategy based on technical indicators. Thereinto, during the time between 7th June,2015 and 1st March,2016, the former’s total profit and yield are 298.38 thousand and 192% respectively, which are far more huge than the latter’s total profit and yield—at 129.72 thousand and 126.57% respectively. What’s more, during the same period, the withdrawal rate of strategy based on basic market indicators is only one third of technical indicators strategy’s withdrawal rate.