開放陸客中轉對國籍航空公司運量變化之研究-以類神經網路建立預測模型

Investigating the market potential of transfer passengers on cross-strait flights: Using the neural network prediction model

蕭仁滔(Ren-tau Hsiao)

指導教授 ： 黃寬丞

國立交通大學/管理學院/管理學院運輸物流學程/碩士(2016年)

摘要

依據IATA預測，2015年至2034年間航空客運能維持每年平均4%之成長，而中國大陸之成長，預估每年將以5.3%之成長率，超越全球成長量之平均值。
2016年2月2日大陸開放南昌、重慶、昆明三點試行，全面開放陸客中轉時間未定，若能全面開放勢必為台灣航空產業帶來巨大之商機。除對航空及旅遊業者帶來利多，對於我國發展區域航空轉運中心有很大助益。
本研究以IATA 2014年Q1至2015年Q3由大陸各航點至北美主要航點之運量資料(含起訖點、中停點、乘載航空公司、票價、旅次、營收)，搭配OAG相關班表、距離及座位數等相關資料，透過類神經網路(Neutral network)之機器學習模型進行建模，來預測2016年全面開放陸客來台中轉後，可吸引之中轉旅客量。
研究結論發現，類神經網路確實可建立預測模式，隨著愈來愈多視覺化大數據分析工具之出現，未來預期將有更多此方面之研究。另外研究發現票價之影響力沒有想像的重要，距離、座位供給數及航班頻率等因素占較重要之影響力，對於不同航線性質要有不同之行銷方式。即使陸客中轉預測之準確性受到政治因素影響大，但此研究結果仍可運用在新增航點之旅運量預測。
本研究之結果可提供航空公司掌握航班規劃、機隊引進、新航點設計及航班銜接規劃，對於機場公司，可預估旅客成長幅度，了解機場作業能量是否能負荷新增中轉旅客運量，及早因應相關配套設施，對於民航單位可了解機場設施之容量及競爭優勢，針對重點來提升機場運作及國際競爭力。

關鍵字

陸客中轉 ； 類神經網路 ； 運量預測 ； 機器學習

並列摘要

By the forecast of IATA (International Air Transport Association), the growth of airlines industry from 2015 to 2034 can remain 4% grow every year. The growth of China is even better and over the average of the industry, 5.3% every year. China may become the largest airlines market in 2032.
Even 13 years pass after the first cross-strait charter flight between Taiwan and Shanghai in January 2003, Mainland Chinese residents are still unable to have a transit flight via Taiwan to other destination due to the political issues. Finally on 02 February, China open three pilot stations (Naichang、Chongqing and Kunming) to free the transit restriction. It will be expected to bring a large amount of passenger stopover via Taiwan, which would not only increase the traffic volume but also help Taiwan develop into a regional air transportation hub.
This research is based on the data from IATA and OAG from 2014 Q1 to 2015 Q3 including the real flight schedule, traffic volumes, frequencies, fare, distance, seats information. And try to develop the prediction model between major origins of China to 4 North America Hub (Las Angeles, San Francisco, Vancouver and New York) using the B-P (Back Propagation) Neural network model.
The results would provide clear picture of the demand of mainland Chinese transit passengers via Taiwan. Airlines can use this information for long term fleets explanation and schedule planning. For airport companies, this study brings them the idea how to deal with the traffic growth as well as airport capacity and facility planning. It also brings the Civil Aviation Authority how to increase the efficiency and competitiveness to attract more transit passengers from Mainland China.

並列關鍵字

Cross-strait flights ； Neural network ； B-P (Back Propagation) ； Market share estimation

參考文獻

王薏婷, et al. (2010). "應用倒傳遞網路預測日本來台航空客運需求之研究." 國立高雄餐旅大學.

郭仲偉 (2008). "以二元羅吉特模式探討大型航空樞紐城市於空間與時間可移轉性之研究." 中國土木水利工程學刊 20(2): 241-256.

黃光浩 (2014). "兩岸航線中轉市場之研究─以台北-洛杉磯航線為例." 交通大學管理學院運輸物流組碩士論文.

Chen, H.-T. and C.-C. Chao (2015). "Airline choice by passengers from Taiwan and China: A case study of outgoing passengers from Kaohsiung International Airport." Journal of Air Transport Management 49: 53-63.

Chen, S.-C., et al. (2012). "Improving the forecasting accuracy of air passenger and air cargo demand: the application of back-propagation neural networks." Transportation Planning and Technology 35(3): 373-392.