굴 소매가격 예측 모형의 DM 검정 및 변동성 분석

Abstract

This study aims at checking the predictive accuracies in reference model and several other models through a process of hypothesis testing by using Diebold & Mariano test (1995) after selecting an ARMA model suitable to forecast through the comparison of errors between predicted value and true value in each model following the prediction of oyster price, using ARMA model. Oyster prices were predicted through ARMA model using daily data. And the predictive accuracies of each model were checked through Diebold & Mariano test (1995) by comparing actual prices with predicted values for the next 1 year. As a result, when both MSE and MAE were considered, it was verified that the predictive accuracies of ARMA(3, 3) model and ARMA(1, 2) model was the most excellent throughout all lags such as 1 day, 5 days, 10 days, 15 days and 20 days. Besides, according to the results of checking volatility through GARCH model, it was verified that variance was persistent for a long period of time once a variance occurred, because lambda value meaning the persistence of variance was close to 1. This study is significant because the predictive accuracies of various models are compared through the process of hypothesis testing by using Diebold & Mariano test (1995), which has not been tried, not by simple comparison of estimates. It is expected that this study will be helpful as a precedent study in checking the predictive accuracies of various forecast models in a study through time series analysis on fisheries area in the future.