Multi scale nonlinear ensemble model for foreign exchange rate prediction

He Kaijian, Xie Chi, Keung Lai Kin

    Research output: Chapters, Conference Papers, Creative and Literary WorksRGC 32 - Refereed conference paper (with host publication)peer-review

    3 Citations (Scopus)

    Abstract

    This paper proposes a novel multi scale nonlinear ensemble methodology for analyzing and modeling the complex exchange rate behaviors. Using several techniques integrated under the proposed unified framework, it deals with data characteristics such as autocorrelation, multi scale heterogeneity and parameter instability during the modeling process. The multi scale heterogeneity property is modeled using wavelet analysis while autocorrelation property is modeled under ARMA framework. Combining independent component analysis, the proposed approach improves the model specification stability using support vector regression based nonlinear ensemble framework. Euro market is chosen as the test case for the performance evaluation of the proposed approach. Empirical studies results suggest that the proposed approach improves the forecasting accuracy and stability. It also offers valuable information as to the underlying micro market structure. © 2008 IEEE.
    Original languageEnglish
    Title of host publicationProceedings - 4th International Conference on Natural Computation, ICNC 2008
    Pages43-47
    Volume7
    DOIs
    Publication statusPublished - 2008
    Event4th International Conference on Natural Computation, ICNC 2008 - Jinan, China
    Duration: 18 Oct 200820 Oct 2008

    Publication series

    Name
    Volume7

    Conference

    Conference4th International Conference on Natural Computation, ICNC 2008
    PlaceChina
    CityJinan
    Period18/10/0820/10/08

    Research Keywords

    • Independent component analysis
    • Nonlinear ensemble
    • Support vector regression
    • Wavelet analysis

    Fingerprint

    Dive into the research topics of 'Multi scale nonlinear ensemble model for foreign exchange rate prediction'. Together they form a unique fingerprint.

    Cite this