Get access

Application of generally weighted moving average method to tracking signal state space model

Authors

  • Cheng-Yi Lin,

    1. Department of Industrial Management, National Taiwan University of Science and Technology, Taipei, Taiwan, Republic of China
    2. Department of Industrial Engineering and Management, Tungnan University, ShenKeng, New Taipei, Taiwan, Republic of China
    Search for more papers by this author
  • Shey-Huei Sheu,

    1. Department of Industrial Management, National Taiwan University of Science and Technology, Taipei, Taiwan, Republic of China
    2. Department of Statistics and Informatics Science, Providence University, Taichung, Taiwan, Republic of China
    Search for more papers by this author
  • Tsung-Shin Hsu,

    1. Department of Industrial Management, National Taiwan University of Science and Technology, Taipei, Taiwan, Republic of China
    Search for more papers by this author
  • Yan-Chun Chen

    1. Department of Industrial Engineering and Management, Tungnan University, ShenKeng, New Taipei, Taiwan, Republic of China
    Search for more papers by this author

Abstract

In predicting time series, if a trend includes a structural break, then a state space model can be applied to revise the predictive method. Some scholars suggest that restricted damped trend models yield excellent prediction results by automatically revising unforeseen structural break factors in the prediction process. Restricted damped trend models add a smoothed error statistic to a local-level model and use the exponentially weighted moving average (EWMA) method to make corrections. This paper applies the generally weighted moving average (GWMA) concept and method to a restricted damped trend model that changes the smoothed error statistic from the EWMA form to the GWMA form and adds the correction parameter λ, which distinguishes three situations math formula, math formula, and math formula. The original restricted damped trend model applies only to math formula, enabling the model to capture situations in which math formula and math formula increases its generality. This paper also compares the effect of various parameter values on the predictive model and finds the range of parameter settings that optimize the model.

Ancillary