최적 시계열 수요예측 모델선정에 관한 연구

Selecting Most Appropriate Time Series Forecasting Model

This study aims to develop four time-series models in order to select the best model among the time-series models based on MAPE(mean absolute percentage error). The time-series models include various ones including ARIMA model. The first three models have been most popularly used for forecasting tourism demand, whereas the last model of ARIMA Intervention is reported to be more logical and accurate than any other time-series models since special events such as terrorism and mega-events can be incorporated into the model. The results of model estimation indicate that all the four forecasting models were found most accurate in terms of MAPE(Lewis, 1982). Of them the ARIMA Intervention model(MAPE=4.48) appeared to perform best in terms of forecasting accuracy, followed by ARIMA(MAPE=4.96), Winters(5.67), and Stepwise Autoregressive(8.55).