The Application Of SVM In The Prediction Of Financial Market

SVM is a method of machine learning according to the statistical learning theory. It is based on VC dimension and structural risk minimization principle. SVM has better generalization and learning power, which has turned into the topic of machine learning, and also gained successful applications in many fields, such as pattern recognition, image classification, forecasting and so on. Up to now, SVM is mainly used to solve classification and regression problems, but rarely applied to time series prediction.Time series forecasting is a subject of technology, which aims to forecast and control through building model by some limited history observed samples and using it to explain the law of the data. The most important function of forecast is to transform the indefinite risk to the estimate one, although the risk is inevitable which can be measured by statistics of probability distribution, such as empirical risk, structural risk and loss function etc. Errors of predicting are inevitable. Predictors have been trying to find better forecasting techniques for reducing the forecast errors, so that the profit of the investors'objective function can be maximized as much as possible and at the same time the risk can be minimized as much as possible. In recent years, neural networks, chaos theory, genetic algorithm, systems theory and contemporary application of the latest developments in mathematics has become a hot topic, as well as and many other theories and methods used in financial time series forecasting of financial engineering. This article with support vector machine method has been applied to forecast financial time series to promote a better application.The introduction of this paper illustrates the value of research and study of the support vector machines. The second chapter details the support vector machine algorithm's basic theory, which is the theoretical basis of the vector classification algorithms in the forth chapter. The third chapter presents the conditional heteroskedasticily model, becoming the base theory of the latter one. In the forth chapter, we use SVM and GARCH model to predict the future trend of financial data. During the experiment, we describe the sources of the experimental data and the way of processing, show the methods and procedures of the experiment, and then predict results: the predictions of SVM is the best.