本研究使用1983年至2006年交通部出版之《中華民國交通統計月報》中的「國人出國按目的地分」月資料,並以香港、日本、韓國、泰國、亞洲區這五個國人主要出國去向為本研究中出國人次研究之範疇。由於過去於灰色預測法之運用上,在許多不同之領域皆得到不錯之運用成果,而近年也運用至國人出、入境之研究中,但主要運用之資料為年資料,雖得到不錯之成果,但若能運用至季、月資料中,相信於各方預測之研究中,會得到更佳之評價。 因此本文主要之方向在於,運用灰色預測法,並對其提出部分修正,使其能有效運用於季、月資料之研究;為了解預測能力之成果是否具有相對優勢,我們同時運作於旅客出、入境研究中得到相當良好預測能力的ARIMA與SARIMA模型,及近年在各方研究中都運用得相當廣泛的類神經網路模型,並加以比較。 結果顯示,經修正後之灰色預測模型,於研究中展現了不遜色於預測能力最好的SARIMA模型之預測能力,並以其少樣本建模與操作簡便之特性,於SARIMA模型不易運用之少數據情況下,也能進行良好的預測。
In earlier studies, grey prediction has been broadly used in different fields. Most of them applied annual data to construct forecasting model. If we use quarterly or monthly data to forecast, the results may be more useful and meaningful. This research amends grey prediction and uses it to forecast. In order to make sure the amended method works well, we then compare the results of the amended grey prediction with other forecasting methods. In this research, we apply grey prediction, ANN, ARIMA and SARIMA to forecast the quarterly and monthly numbers of people going abroad to Hong Kong, Japan, Korea, Thailand and Asia. We found that the performance of amended grey prediction is as good as SARIMA model, which is the best forecasting method in earlier papers. Because of the characters of easy-handled and requiring less data series, the amended grey prediction can also work well in some condition that SARIMA model can’t do.