Financial Time Series Researching Based On Support Vector Machine

Financial market is the essential economic system of a country. And financial time series is a primary data type in the application of financial area. Analyzing, predicting and controlling of such kind of data is the basic work of the economic and financial activity. Financial time series forecasting is regarded as one of the most challenging applications of modern time series forecasting because of its characteristics of nonlinear and the small sample.Statistical Learning Theory(SLT) focuses on the learning theory of small samples. The core of the theory is to control the generalization of learning machine by controlling the complexity of models. Support Vector Machine(SVM) is a general learning algorithm developed from SLT. It has been successfully used in pattern recognition, regression and time series prediction. Support Vector Regression(SVR) is the expansion of SVM to regression problems.Decision tree, is one of the most widely used and practical methods for inductive inference because of easily understandable and high classification accuracy. It can perform automatic feature selection and complexity reduction. In this Paper, we have constructed a SVR model which is based on decision tree algorithm for feature selection task of financial time series. Our experiment results show that the combination of the decision tree and SVR leads to a better performance.SVM uses kernel function to extend to nonlinear problems by using its the special nonlinear mapping for feature space. To choose or construct appropriate kernel for a given problem is important to improve the performance of SVM. In this paper, we have constructed a SVM with a mixture kernel using polynomial kernel and Radial Basis Function(RBF) kernel under the instruction of Mercer theory. Then we have applied the mixture kernel SVM in financial time series forecasting. Experiments show that SVR with mixture kernel has better performance than which with the individual kernel.