Research On High Frequency Quantitative Investment Based On Support Vector Machine

Quantitative investment has a history of more than thirty years in the oversea market and its market occupancy is still increasing because of its stable modeling investment performance. The essence of the quantitative investment is modeling trade. The quantification process is carried out to model the trading behavior, make the parameters of trading variables quantitative, once market conditions being triggered,trading actions will be executed.The theory of support vector machine is first proposed by Corinna Cortes and Vapnik in 1995. It is a kind of machine study algorithm that based on statistic theory and the basis of support vector machine is a solid statistical theory that it does not converge to the local optimal solution, making it’s distinguish from other machine learning algorithms.Considering the advantage that support vector machine is suitable to solve the problem of nonlinear data and small sample. It also has many unique advantages in solving high dimension pattern recognition problems. This paper considers the regression prediction of IF data in stock index futures, and studies from the following aspects:On the one hand, in view of the choice of kernel function and even the construction of kernel function has no adaptable theoretical guidance so far, support vector machine is sensitive to the parameters. This paper selects the RBF kernel function as studying sample, focusing on the optimization of the g parameter of kernel function and penalty factor C,the parameters were optimized by improved grid method, parameter optimization effect is excellent.In addition to the importance of the high frequency trading quantitative model, the transaction costs are important factors affecting the results, especially the bid-ask price spread and the fee level. So the analysis of the level of the bid-ask price spread and fee level is essential when the model was built. This paper uses logarithmic linear regression method on spread of fitting for cost accounting model, getting a significant effect. Cost model is figured out at last when the strategy is achieved by devaluing the bid-ask spread cost.In this paper, fitting and classification prediction model is built and it studies the feasibility of the strategy construction of the high frequency quantitative investment in the domestic market based on IF stock index futures contract, and conducts the practical work on the basis of the research process. The strategy is achieved based on the spread analysis model and the support vector machine prediction model,showing the threshold of the fees in different markets. Theoretical analysis is verified by actual effect, has practical significance.