航空運量之預測，自策略規劃至實際營運皆具不可忽略之重要性，本研究採取監督式學習網路中之倒傳遞神經網路(Back-Propagation Network)，自行撰寫MATLAB程式，建構日本來台航空客運需求預測模式。本研究透過回顧相關文獻，市場分析與初步評估，決定輸入變數，再運用去除法、加入法與貢獻圖，檢驗其影響程度。實證結果發現以日本人口、日本就業人口、日本個人所得、日本GDP、日本GNP、外幣匯率、航班數量、臺灣個人所得等8項變數，預測日本來台航空客運量之平均絕對百分誤差(Mean Absolute Percentage Error)爲0.34%，預測能力極佳。探討輸入變數得知，航班數量、臺灣個人所得與外幣匯率之影響最大。

Air traffic demand forecasting plays the important role for developing the efficient operating strategy. This study employed supervised back-propagation neural networks to develop a simple MATLAB computer program in order to forecast air passenger demand from Japan to Taiwan. All input variables were collected by literature survey, market analysis and preliminary evaluation. The factors which influence air passenger market were analyzed in detail by deleting method, adding method, and contribution graph. The following 8 variables were selected as the input variables to establish the forecasting model: population in Japan, employed population in Japan, per capita income in Japan, GDP in Japan, GNP in Japan, foreign exchange rate, flight movement from Tokyo (NRT) to Taipei (TPE), per capita income in Taiwan. The novel BPN model can accurately forecast air passenger demand with an extremely low Mean Absolute Percentage Error of 0.34%. The results reveal that flights from Tokyo (NRT) to Taipei (TPE), PCI in Taiwan, and foreign exchange rate are the three most important factors for air passenger volume.