摘 要 觀光需求預測長久以來是觀光遊憩規劃中重要的議題。關於預測方法的運用，據 統計達300多種，運用於觀光需求或遊客人次預測之量化評估則多以ARIMA 預測模 型、迴歸分析、重力模型等方式進行實證研究。然而由於上述方法需要大量的資料， 往往無法面對環境內外部因子的快速變化。灰色系統理論的產生則可以提供不確定環 境灰色因子的建模，透過灰色生成、GM 模型建構與誤差檢測，提供觀光產業預測的 新方向。本研究以國家風景區作為實證範圍，透過整併方式，對遊客人次以全國、區 域與個別國家風景區的層次進行加總，並進行遊客人次預測。結果顯示灰色系統理 論，在各層次所提供的預測值，皆較過去建構之預測模型誤差值更低。且發現透過不 同層次劃分，顯示較精細之範圍劃分，所獲得之預測值合併，較經整併後之預測值精 確度高。研究結果引導出灰預測模型應用的限制，亦提供經過不同整併方式可提高精 確度之建議。

ABSTRACT The prediction of tourism demand has been an important factor as tourism planning is proceeded. According to some references, the number of prediction methods is more than 300, and these methods used in the estimation of tourism demand are involved with ARIMA, Regression Analysis, Gravity Mode. However, these methods can not afford to deal with the swift transformation of the inner and outer factors, because they need a lot of data. The appearance of Grey System Theory may provide a new direction to solve these problems,and it can build a model to predict the tendency of tourism demand through Grey Generating and error examination under the unreliable environment. This research try to arrange Taiwanese National Scenic Area in different divisionsn and predict the number of the tourists in National Scenic Area. The result shows that this theory can provide more accurate prediction in any divisions. Furthermore, it also proves that the accuracy can be improved through the predictions of more detailed units rather than an abstract division. As a result, this paper implies the restriction of grey predicting theory and provides more accurate way in predicting the number of tourists with the method.