本研究進行日本及美國兩個已開發國家來台灣觀光旅遊人次之時間序列分析。本研究以月資料和季資料，使用四種ARIMA 時間序列延伸模型進行模擬並檢測各模型預測力。研究結果顯示，SARIMA 模型能含納觀光旅遊人次系統性季節變化，因此有最佳預測力。研究亦發現月資料能保有較多遊客人次時序變化資訊，使用月資料進行分析優於季資料。模擬結果發現日美來台人次在1991 到2009 年研究期間為定態序列，並未有明顯增減趨勢，顯見此期間施行之觀光客倍增計畫並未形成日本與美國來台人次長時間增加的效果。以預測力最佳之SARIMA 模型模擬預測未來短期來台人次，結果亦僅呈微幅成長現象。

This paper analyzed the time series data of tourism arrivals of Japan and America by using monthly and seasonally data with four extended ARIMA models. Monthly data are evidenced much powerful than the seasonally data in simulation and forecast. SARIMA model is evidenced the superior model in the ability of co-operating systematical data fluctuation in simulation and forecast among the four models. The stationary results of these two time series suggested that tourism arrivals from Japan and American are both stable. Forecasted by the superior SARIMA model, a slight increase in the number of arrivals from Japan and American was suggested in the nearby future.