The Application Of Net Value Of Security Investment Fund Based On Support Vector Regression

Support vector machine (SVM) is one kind of machine learning method based on statistic learning theory. As the implementation of the structure risk minimization, SVM has many advantages, such as global optimization, simple structure, strong extension ability etc. As SVM has strict theory basis and it can solve many practical problems, it becomes one of the most important achievements during the past decades.During the past two decades, investment fund develops quickly in the global Financial Market. Especially during the past several years, the focus of people on fund develops fast. We can see the income through net value of security investment fund and it becomes the main reference of that we buy fund and which fund or not. Net value of security investment fund has very important application value. The author construct a predict model of net value of security investment fund based on support vector regression and it makes very good effect. This paper firstly studies the theoretical basis of securities markets predict and the present condition of stock and fund. The predict methods of fund are mainly qualitative; though some use regressive equation analysis methods to predict fund, but all of them are lack of predict result of quantitative. So it indicates the necessity and usefulness to use support vector regression to study security investment fund. Secondly, the author mainly introduces statically learning theory and some kernel concepts of SVM and support vector regression algorithm. Secondly, the author mainly introduces statically learning theory and some kernel concepts of SVM and support vector regression algorithm. A single parameter relaxation two norm support vector regression is deduced based on standard support vector regression. Finally, the author realizes the predictions of two net values of funds by the means of MATLAB programming and one step of BP network iterative. And then, nonlinear time series prediction model of SVM is constructed to predict the two net values of funds. The two results are compared and analyzed. Simulation tests show that prediction model can not only forecast the net value of fund well, but also acquires good effect in nonlinear prediction. So it has a certain application value.