시계열 분석방법을 이용한 과채류 월별가격 예측

Abstract

The purpose of this paper is to conduct the short-term price forecasting for six varieties, three Fruit-Vegetable items and construct the adequate model for each item by comparing forecasted results with actual results. The analysis data is the monthly Garak-dong wholesale market price from January, 1995 to December, 2005. The period of price forecasting is from January, 2006 to June. Models for the analysis are considered as to be VAR, ARIMA, GARCH model. Considering the substitution of production, this study increases multi-variate analysis of time-series data and derive some findings by comparing multi-variate analysis with uni-variate analysis. The findings are given below.1.ARIMA model has higher accuracy of price forecasting than VAR model except Young pumpkin which is only 6.2 percents of forecasting error in VAR model. To explain, in VAR model, ranged is forecasting error from 10~20 percents(Bacdadagi cucumber, Big tomato, Mini tomato) to more than 20percents(Jukini pumpkin). Unlike this, in ARIMA model, Bacdadagi cucumber and Young pumpkin are less than 10 percents and the others are 10~20 percents.2.The result of price forecasting through ARMA-GARCH model for Young pumpkin which has ARCH effect is better than ARMA model.3.In all models, the best accuracy of price forecasting is Young pumpkin and Bacdadagi cucumber, Big tomato, Mini tomato are normal. However, Chichung cucumber and Jukini pumpkin are only good in ARIMA, VAR model in the order.Conclusively, suggested is that this paper establishes appropriate time-series models for each model by using analysis techniques of time-series. Therefore, this paper can be used as useful data to forecast Fruit-vegetable price for the related agricultural people.