Dissertation > Excellent graduate degree dissertation topics show

Modeling and Forecasting Seasonal Time Series Using Seasonal Support Vector Regression

Author: QianJiFu
Tutor: YangJianHui
School: South China University of Technology
Course: Quantitative Economics
Keywords: Seasonal support vector regression Exponential smoothing SARIMA Seasonal time series Forecast
CLC: F224.0
Type: Master's thesis
Year: 2010
Downloads: 180
Quote: 1
Read: Download Dissertation

Abstract

Support vector machines (Support Vector Machine, SVM) is a new machine learning algorithms based on statistical learning theory, the idea of ??structural risk minimization, SVM has better than traditional methods based on empirical risk minimization criteria learning performance and generalization performance, and with the complex problem of the small sample, nonlinear modeling capabilities. Support vector regression (Support Vector Regression, SVR) since Vapnik in the definition of the ε-insensitive loss function based on support vector machine algorithm for regression forecast, in theory, have the advantage of the same SVM, SVR since the introduction of aroused wide interest of researchers has been successfully applied in many fields, but common seasonal time series for the areas of business, economic, literature and there is no specific research. Construction and application of this study SVR seasonal time series forecasting model focuses on the design of data preprocessing on the performance of the SVR prediction modeling method. Specifically, the chapter is organized as follows: first of all through the issues raised, the literature review found that the lack of existing research, asked the purpose of this study, research, and possible innovations; then through a review of the traditional seasonal time series forecasting methods - exponential smoothing and SARIMA, lay the foundation for the later to assess seasonal SVR forecast performance; Chapter three seasonal SVR model modeling ideas (to use the raw data Raw Data-SVR with a differential of season pretreatment strategies season differential-SVR, and a combination of the seasonal adjustment method and the corresponding data change operation Seasonal-SVR) and describes the modeling steps; Chapter IV of this application study chapters, macro and micro, different size sample data empirical test seasonal SVR predictive ability of the model as well as its comparison with the traditional model; Finally, we come to the following conclusions: (1) the use of the original data SVR seasonal time series prediction is not ideal; ⑵ season sex time series suitable pretreatment, can significantly improve the prediction performance of the SVR; ⑶ In this paper, two data preprocessing strategy: the the season season differential-SVR differential Seasonal-SVR in the moving average ratio the multiplicative model combined with the corresponding data transform operation, the empirical results show that the latter than the former; ⑷ season differential-SVR and Seasonal-SVR, the prediction accuracy have small sample seasonal time series modeling and forecasting capabilities to smooth seasonal model and SARIMA; traditional index ⑸ the SVR In this paper, two modeling methods seasons differential-SVR and the Seasonal-SVR prediction accuracy are higher.

Related Dissertations

CLC: > Economic > Economic planning and management > Economic calculation, economic and mathematical methods > Economic and mathematical methods > Quantitative Economics
© 2012 www.DissertationTopic.Net Mobile