Journal of Rural Development
Forecasting Spice Vegetable Prices Using Forecast Combinations and Assessment of Forecasting Performance
-
AuthorHan, Eunsu
-
Publication Date2025.09.21
-
Original
This study aims to determine whether forecast combinations improve predictive accuracy compared to individual price forecasting models. This study estimated forecasting models, including ETS, ARIMA, and Artificial Neural Networks, for three major spice vegetables: onion, garlic, and dried red pepper. The analysis evaluated whether combining forecasts from individual models improved predictive accuracy compared to using each model separately. The forecast combination methods employed included simple averaging and median-based combinations, regression-based combinations, performance-based combinations using Mean Squared Error (MSE) and MSE rank, and combinations based on the Akaike Information Criterion (AIC). The results showed that combining forecasts from individual models improved predictive accuracy for all three spice vegetables. Moreover, as the forecast horizon extended from 1 to 6 months, the prediction errors of the combined forecast were smaller than those of the individual models, highlighting the effectiveness of the combination approach. Among the methods, performance-based approaches using MSE and MSE ranks yielded the best forecasting performance. This study is academically significant because, unlike previous domestic studies that mainly focused on evaluating individual models, it applied forecast combination methods to agricultural price forecasting in Korea.
- Next
- Development of a Domestic Pork Supply and Demand Forecasting Model Using a Simultaneous Equation Model
- Prev
- A Study on the Change of Issues Related to Companion Animal Using Keyword Network Analysis