Financial Time Series Researching Based on Support Vector Machine

Financial markets is the core of the national economic operation , financial time series is the most important in the field of economic and financial data type , the type of data analysis , forecast and control the important work of the entire economic and financial activities . Financial time series due to its non-linear and small sample characteristics , become the most challenging topics in the study of modern time series . Statistical Learning Theory (Statistical Learning Theory, SLT) for machine learning theory in the case of the small sample size , the core idea of ??the generalization ability of the learning machine control is achieved by controlling the degree of complexity of the learning machine . Developed this theory support vector machines (Support Vector Machine, SVM) is a new general-purpose learning methods , showed on the theoretical and practical advantages compared to the conventional method , already in pattern recognition , regression estimation , time sequence prediction has been successfully applied in many ways . Support vector regression (Support Vector Regression, SVR) and support vector machine is used to solve the generalized form of regression problems . Decision tree algorithm is simple and high classification accuracy has become a widely used method of inductive reasoning , can complete the complexity dimensionality reduction and automatic feature extraction . Paper , we have created a financial time series - based decision tree characteristics extracted SVR regression model . The experiments show that the integration method based on decision tree and support vector machine can effectively improve the performance . SVM feature space via the kernel function mapping function to achieve the promotion of non-linear case . Select or construct specific problem for the kernel function is an important way to improve the SVR performance . Mercer kernel function tectonics theory , under the guidance of the constructed based on polynomial kernel and Gaussian radial basis ( Gaussian Radial Basis Funciton RBF) nuclear new the mixed kernel support vector machine model . The results show that a better performance than the single -core support vector machine based on the mixed kernel support vector machine in financial time series prediction .