EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AN ENHANCED REGIME-SWITCHING MIDAS MODEL WITH AN AUTOREGRESSIVE COMPONENT, HIDDEN MARKOV MODELS, AND SELF-ORGANIZING MAPS FOR FORECASTING MIXED-FREQUENCY TIME SERIES

Phakkhaphon Sawatkamon, Eckart Schulz, Jessada Tanthanuch

Far East Journal of Mathematical Sciences (FJMS) · 12.08.2026 · 143 (10) · 3423-3439

DOI: 10.17654/0972087126185 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Far East Journal of Mathematical Sciences (FJMS), online erschienen am 12.08.2026. Verfasst von Phakkhaphon Sawatkamon und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study proposes an enhanced regime-switching Mixed Data Sampling (MIDAS) model, denoted as MIDAS-AR-HMM-SOM, by integrating an autoregressive (AR) component, Hidden Markov Models (HMMs), and Self-Organizing Maps (SOMs) to improve forecasting accuracy for mixed-frequency time series under changing market conditions. The forecasting performance of the proposed MIDAS-AR-HMM-SOM model is compared with three benchmark models, namely, MIDAS, MIDAS-AR, and MIDAS-AR-HMM, using monthly gold prices as the target variable and weekly crude oil prices together with the U.S. Dollar Index (DXY) as high-frequency explanatory variables over the period from 2001 to 2025. Model performance is evaluated using the Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Square Error (RMSE) under three training-to-testing ratios of 70:30, 75:25, and 80:20. The results demonstrate that the proposed MIDAS-AR-HMM-SOM model consistently outperforms the benchmark models across all evaluation settings. Under the 70:30 training-to-testing ratio, it achieves the best forecasting performance with an MAE of 43.973, a MAPE of 2.133%, and an RMSE of 61.899. Furthermore, pairwise Diebold-Mariano (DM) tests indicate that the improvements over all benchmark models are statistically significant, confirming the robustness of the proposed forecasting framework. These findings demonstrate that integrating HMM-based regime identification with SOM-based nonlinear pattern learning effectively enhances the predictive capability of the MIDAS framework for mixed-frequency time series forecasting.

Zitieren

APA

Sawatkamon, P., Schulz, E., & Tanthanuch, J. (2026). AN ENHANCED REGIME-SWITCHING MIDAS MODEL WITH AN AUTOREGRESSIVE COMPONENT, HIDDEN MARKOV MODELS, AND SELF-ORGANIZING MAPS FOR FORECASTING MIXED-FREQUENCY TIME SERIES. Far East Journal of Mathematical Sciences (FJMS), 143(10), 3423-3439. https://doi.org/10.17654/0972087126185

Vancouver

Sawatkamon P, Schulz E, Tanthanuch J. AN ENHANCED REGIME-SWITCHING MIDAS MODEL WITH AN AUTOREGRESSIVE COMPONENT, HIDDEN MARKOV MODELS, AND SELF-ORGANIZING MAPS FOR FORECASTING MIXED-FREQUENCY TIME SERIES. Far East Journal of Mathematical Sciences (FJMS). 2026;143(10):3423-3439. doi:10.17654/0972087126185

BibTeX
@article{Sawatkamon2026,
  author = {Sawatkamon, Phakkhaphon and Schulz, Eckart and Tanthanuch, Jessada},
  title = {AN ENHANCED REGIME-SWITCHING MIDAS MODEL WITH AN AUTOREGRESSIVE COMPONENT, HIDDEN MARKOV MODELS, AND SELF-ORGANIZING MAPS FOR FORECASTING MIXED-FREQUENCY TIME SERIES},
  journal = {Far East Journal of Mathematical Sciences (FJMS)},
  year = {2026},
  volume = {143},
  number = {10},
  pages = {3423--3439},
  doi = {10.17654/0972087126185}
}
Kostenlos starten →