The Application Of SVR In The Prediction Of Financial Time Series

Classical statistical theory makes great contributions to the treatment of low dimensional data classification and estimation problem. In the 1990 s the American Vapnik had put forward Support Vector Machine(Support Vector Machine, SVM) theory, it is a new system for the small sample, it not only consider the gradual problem, but also get the optimal solution of the problem with the limited conditions. At present the support vector machine is mainly used in application and research of classification and regression problems; however it also has the very big promotion space in the prediction aspect. In the actual financial data, the time series data is mutually dependent and interrelated, the development of financial markets in the past and the present, will directly affect the development level of financial market in the future. Using time series analysis theory to build the reasonable models, it can rely on the historical data sample to predict unknown data, then we can speculate the future development trend of financial markets and its value at risk, as far as possible reduce unnecessary economic losses. In actual application, building time series model is only an approximation approach to the actual data, rather than completely identical, so prediction of unknown data need to be proven. In general, different models are used respectively to model and predict, then choice whose accuracy and precision is higher.In this paper, firstly we introduce a few time series analysis theories briefly, and respectively enumerate several time series models and conditional heteroskedastic models, then based on this, we more introduces the principle, algorithm and achieve process of the Support Vector Regression in detail. And use these methods to daily closing price of southwest securities and the Shanghai composite index to prove. SVR is also compared with traditional time series prediction method, and the former perform evidently is better than the latter’s, which reflects the support vector regression in prediction of financial time series data to have good effectiveness.