The Net Value Of The Fund Prediction Model Based On The Wavelet Denoising And SVM

During the past two decades,investment fund debelops quickly in the globalFinancial Market.We can see the income through net value of security investmentfund and it becomes the main reference of that we buy fund and which fund ornot.Net value of security investment fund has very important application value.Support vector machine(SVM) is one kind of machine learning method based onstatistic learning theory. As the implementation of the structure risk minimization,SVM has many advantages, such as global optimization, simple structure,strong extension ability etc. It can solve many practical problems, it becomes oneof the most important achievements during the past decades.However,the data in the real world has so many points of largeamount,incomplete and noisy.,which Influences of the accuracy of prediction model.In the prediction of the fund net worth, we are most concerned about is embedded inthe daily fund net net change in trend,but not those of man-made or daily fund netmarket fluctuations caused by noise or burr.In view of the above two key problems, this paper studied the model based onthe wavelet denoising and SVM. This paper has completed the following jobs:(1) This paper discusses that using wavelet denoising methods on data pretreatment,and puts forward the improvement of the wavelet threshold denoising method. Itindicates the new method has better demonstrate with the Matlab toolkit.(2) In this part,a new kernel named financial kernel had been constructed for theimprovement of the prediction accuracy, the relevance parameters selection andcontrol are also investigated.It indicates the new kernel function has better demonstrate with the LibSVM toolkit.(3) This paper founds the net value of the fund prediction model based on thewavelet denoising and SVM.It indicates the new model has better demonstrate withthe two funds,comparing with BP neural network prediction.