Journal of Rural Development
Analysis of Forecasting Performance of a Kimchi Cabbage Price Forecasting Model Using Time Series Decomposition and Ensemble Methods
-
AuthorHan, Eunsu
-
Publication Date2026.09.21
-
Original
This study proposes a decomposition-ensemble algorithm that combines time series decomposition with forecast combinations to improve the short-term forecasting performance of monthly wholesale kimchi cabbage prices. The log-transformed price series was decomposed into seasonal and seasonally adjusted components using STL decomposition. The seasonal component was forecast using the seasonal naive method, whereas the seasonally adjusted component was forecast using ETS, ARIMA, ARIMAX, and neural network autoregression (NNAR) models. The resulting forecasts were combined using the simple average, median, and MSE-based weighted average. The empirical results suggested that the proposed decomposition-ensemble model achieves lower forecast errors than models estimated directly from the original series and individual decomposition-based models. In addition, a comparison with the KREI Agricultural Outlook Center’s price outlook and an encompassing test indicated that the two forecasts contain complementary information. These findings suggest that the proposed algorithm can serve as a supplementary quantitative tool in the Agricultural Outlook Center's forecasting system.
- Next
- Asymmetric Exchange Rate Pass-Through to Retail Meat Prices: A NARDL Approach
- Prev
- The Effects of Tourism Motivation and Destination Selection Attributes on Actual Revisit to Rural Tourism Destinations