A Study Of Quantitative Timing Based On PCA-SVM Model

With the development of Chinese stock market, the function of this market is increasingly high. The research of stock market is much more than before. This paper studies about HS300 index with Technical indicators and the Support Vector Machine (SVM) theory. In the paper, profitability and stability of the investment are the main point, which utilize the statistical software to make quantitative timing empirical analysis. In addition, HS300 stock index futures have launched in China, this research can contribute to the investment strategy.The ability of the traditional technical analysis and fundamental analysis is much weaker currently, so this paper analysis Chinese stock market price volatility using the nonlinear data mining technology. In the comparison of technology choice, finally, the support vector machine is chosen because of the higher accuracy than artificial neural network. Getting deeper research in SVM, there are two stages of SVM, namely traditional and improved ones. The improved support vector machine is good at the optimization of parameters,, and the optimization grid search method in empirical accuracy performance was higher.Based on the quantitative timing theory and ADF unit root test to the data, it is determined that this part of data has reached weak form efficient. In processing of data, the Principal Component Analysis (PCA) is utilized in order to reduce the miscellaneous and improve the accuracy of the model. The empirical result shows that PCA has a better accuracy.In according to this, the paper constructs the PCA-SVM model with SVM, PCA, and grid search method. Combining the data and quantitative timing method of SVM model, and compared to the HS300 index over the same period, it is concluded that SVM has the higher profitability, stability and con-dropping ability.However, for the reason of lack in knowledge and experience,this paper could not merge the technical indicators and macroeconomic data that reduce the interpretation ability.And this paper is not concern about the noise that exists in the stock market, which is not just the correlation in the data. These aspects need further research.