Font Size: a A A

Study On The Choatic Time Series Prediction Method Based On Support Vector Machines

Posted on:2007-12-24Degree:MasterType:Thesis
Country:ChinaCandidate:J L DangFull Text:PDF
GTID:2178360182961722Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
With the development of chaos theory and application technology study, analysis and prediction of chaotic time series have become the emphasis of chaotic signal precessing research domain, and can solve a lot of nonlinear signal process questions in engineering practice, which are difficult to be done by linear signal processing methods.With their introduction in 1995, Support Vector Machines (SVM) marked the beginning of a new era in the learning from examples paradigm. Rooted in the Statistical Learning Theory (SLT) developed by Vapnik and co-workers at AT&T Bell Laboratories, SVM has been a hot research issue in the international machine studying field and quickly gained attention from the pattern recognition community due to a number of theoretical and computational merits. These include, for example, Strict theory ,the simple geometrical interpretation of the margin, unique and globally optimal solution, good generalization performance, statistical robustness of the loss function, modularity of the kernel function, and overfit control through the choice of a single regularization parameter. Owing to its stabile basis in theory and excellent studying features, combining Support Vector Machines with chaotic theory and classic chaotic prediction method, the prediction approach based SVM been discussed in this paper, which show that the application is successful and meaningful, therefore ,with solving the real world problem ,we extend the field of SVM applications..Research works focus on SVM and its application in prediction of chaotic time series in this dissertation, which mainly include analysis of chaotic time series and how to select the appropriate values for the parameters in the global and local SVM prediction of chaotic time series, local prediction of chaotic time series using support vector machine is proposed to make prediction of spatiotemporal chaotic time signals. The main research fruits are as follows:1 , At first, chaotic time series prediction based on SVM is expatiated in this paper, then the feasibility and prediction performance of chaotic time series with SVM method are studied completely. Tuning the parameters of the system is another important design issue in the SVM, so the generalization error with respect to the free parameters of SVM is investigated in this paper.2, Based on the deterministic and nonlinear characterization of the chaoticsignals, combined the advantage of traditional local linear prediction with remarkable characteristics of SVM. The forecasting model of support vector machines in combination with chaotic local prediction methods has been established, using this model relations between the embedding dimension and mean-error of this model are discussed, the new regression predict model constructed is applied into predict chaotic time series successfully, the training speed of SVM is largely improved and the inner memory requirement of training SVM is greatly decreased therefore, the methods is feasible and effective, and adapts. to real engineering application. Moreover nonlinear prediction structure and algorithm of chaotic time series are developed further.3. The local SVM prediction method is used to predict three typical spatiotemporal chaotic time signals and global SVM method to chaotic FH code based on the chaotic characteristics analysis of them. Compared with the approach of classic prediction methods, the proposed method has the features of high learning speed, good generalization and better prediction precision without time delay. Analysis of the experimental results demonstrates that SVM prediction methods can predict the chaotic time series efficiently, and are more accurate than adaptive linear prediction method. It is denoted that the SVM and its kernel method has advantage and its feasibility in the prediction of chaotic time series.
Keywords/Search Tags:chaotic time series, support vector machines, phase-space reconstruction, spatiotemporal chaotic time series
PDF Full Text Request
Related items