A1489
Title: An extension of the dynamic seasonal grey model with covariates
Authors: Jenyu Lee - Fengchia University (Taiwan) [presenting]
Abstract: The aim is to extend the dynamic seasonal grey model by generalizing the univariate DSGM(1,1) to a covariate-inclusive DSGM(1,p+1), aiming to enhance forecasting performance for data exhibiting both exogenous influences and dynamic seasonality. In the DSGM(1,p+1) framework, the dynamic seasonal adjustment factors (DSAF) for the forecasting period are estimated using several methods, including the naive approach, HoltWinters, SARIMA, and Fourier expansion. These estimates are then integrated with the predicted values of the underlying trend component to obtain improved forecasting results. Furthermore, various simulation scenarios are designed, considering patterns such as stable seasonality, varying seasonal amplitude, and unstable seasonal structures, along with different types of covariates, to compare the performance of the proposed models. Finally, empirical validation is conducted using real-world data with pronounced seasonal patterns and available covariate information. It is expected that the approach will broaden the theoretical and practical scope of dynamic seasonal grey models and provide an integrated forecasting approach that simultaneously accounts for covariate effects and the estimation of seasonal factors during the forecasting period.
