Based On The Support Vector Regression(SVR) And SV-TT Model Of Quantitative Strategy Research

Quantitative investment--a kind of transaction methods which is based on statistic,considered data model as main core and programming trading as measure.Its features including huge trading volume,short holding time of position as well as stable total revenue.In contrast with traditional investment strategy,quantitative investment has lots upsides.Firstly,it investment logic is more objective--producing and conducting decision by computer,which removes the impression made by humans subjective bias.Whatever,it shows upper efficiency and accuracy.The characteristic makes it easy to keep step with market changes and catch the key information.China during the initial stage in quantitative investment area(only occupies 5 percentage in financial investment),revealing a expansive prospects of quantitative investment in China's stock domain.Along with the development of China's equity market as well as the reform and innovation in share area,quantitative investment is destined to act an important role in future domestic market.This article is based on both disturbance to rush fat-tailed T distribution,on the basis of building can more accurately describe the financial yield fluctuation when degeneration,the aggregation of SV-TT model.Based on the prior distribution of the traditional assumption,the detailed process of SV-TT MCMC estimates of the model was deduced,and use it to the empirical analysis of HS300 index.By compared with the traditional rule of SV model clusters of DIC,confirmed the validity of the SV-TT of volatility forecasting model and superiority.Respectively into the support vector machine(SVM)and SV-TT model to predict the HS300 index r of yield and yield fluctuation,a combination of established in this paper,the quantitative investment strategies.First of all,through the selection of six categories factor,principal component analysis pretreatment,take its first six principal components for input,to each stock after five days of cumulative returns as output.Support vector machine(SVM)regression model is established to forecast for warehouse stocks yield of 5 r in the future.Then,three months prior to the warehouse every five day yield data,SV-TT model,to predict the stock in the warehouse,five days after the yield fluctuation risk.SV-TT models to predict the yield on the standard deviation to modify the predicted yield of support vector machine(SVM)--r,elected in the HS300 stock market(including risk aversion factor)in the pool,and in the candidate pool selected 50 stocks to buy sets of maximum value,every 5 positions.According to the quantitative strategies on HS300 stock market back to measuring results,the annualized yield can be over 33%,total yield reached 117.8%,and the sharpe ratio is 0.83.In the case of bare more,the return on capital is far more than the csi 300 index,shows the superiority of strategy.