- 學位論文
應用深度學習自編碼於臺鐵國定假日之運量分析
Deep Auto-encoder for Analyzing the Ridership of Taiwan Railways Administration on National Holidays
摘要
臺灣除擁有周休二日的休假制度外，每年也另有八個國定假日，依序為元旦、春節、二二八和平紀念日、清明節、勞動節、端午節、中秋節、及國慶日。除了春節連假長達六天(以上)以及少數一天的國定假日外，其餘假期長度多為三至四天。過往由於資料取得問題，針對臺鐵國定假日的運量，並無系統性的研究，近期在取得臺鐵2014至2018年間的所有票證資料後，因而獲得了完整的運量資訊。而進一步比較國定假日與一般週末之運量後，發現因國定假日時間長，其運量的波動程度相對較大，且其旅次需求量多，營運調度相對複雜，如能明確掌握運量分佈之情形，將可提供更完善的應對策略。過往關於運量的研究，大多是針對各種交通運具，如公車、軌道系統(地鐵、火車、高鐵)、航空的運量，進行直接預測；或針對影響運量的因子進行分析。相較之下，針對國定假日的運量特性進行分析之研究寥寥可數，因此本研究期臺鐵對於往後的國定假日，以研究成果為依據，能在事前瞭解每一日可能的運量分佈，將之作為營運調度的參考。 本研究的研究方法為依據每一放假日的本質，用其相對應的「起訖縣市(OD)-日運量分佈」表格，逐日分析其運量特性，使用深度學習自編碼(Deep Auto-encoder)降維、擷取特徵，再使用K-平均演算法(K-means Clustering)將相似的放假日分群，並找出各群的代表特徵，最後建立多項邏輯斯迴歸(Multinomial Logistics Regression)模型，提供明確的歸納彙整。研究結果分為四群，分別是「返鄉返回工作群」、「次返鄉返回工作群」、「次區域旅次群」、「區域旅次群」。「返鄉返回工作群」在城際旅次(中長程)之運量占比則近三成(27.89%)，居四群之冠。「區域旅次群」則擁有超過80%區域型旅次的運量占比，近似於平常日的運量分佈。而「次返鄉返回工作群」、「次區域旅次群」則介於兩者之間。
並列摘要
Besides two days off a week, there are eight national holidays as New Year ’s Day, Lunar year holidays, 228 Peace Memorial day, Ching Ming Festival, Labor Day, Dragon Boat Festival, Moon Festival and Double Tenth Day in sequence a year in Taiwan. Chinese New Year’s holidays are national six days or above. The rest have, most likely, three or four days off when to be coupled with the weekend. In the past, lacking of the research in ridership of the national holidays due to the difficulty in getting the relative data for analysis. Recently after obtaining all the ticketing data, to analyze and compare the traffic volume between weekend and the national holidays gains a conclusion as follows. The daily ridership for the national holidays is with more fluctuations comparing to the weekend. Obviously, to grab, accurately, the distribution of ridership is great help to find out the proper strategy. Hence, this article focuses on exploring the ridership of the national holidays. The previous research, mostly, aims at directly predicting the ridership for the various carriers of transportation such as bus, railway system including subway, train and high-speed rail and air transport. Or, analyze the factors which will affect the ridership. However, the objective of this research is to let the Taiwan Railway Administration can realize the possible ridership distribution of national holidays in advance and take action based on our outputs. This article is using the ridership distribution table from Original County to Destination County, and utilizing a deep learning auto-encoder by dimensionality reduction method to extract the feature. Then, re-utilize K-means Clustering to have the similar holiday grouped and find out the feature of each group so as to establish “Multinomial Logistics Regression” model to provide the clear and proper induction. Then About the group one, the ridership which belong to the Regional trip occupies 63.96% of the average daily ridership. The ridership of Intercity Trip takes 27.89% (approximate to 30%), which is the greatest of the 4 groups. The fourth group owns exceeding 80% of ridership of Regional trip. It’s similar to the distribution of travel volume of weekday. The second and third groups are distributed between the first and the fourth groups.
參考文獻
延伸閱讀
- Chang, M. Y. (2014). 整合及運用資料包絡分析法與平衡計分卡對於績效之研究以臺灣地區金控公司為例 [doctoral dissertation, Chang Jung Chrisian University]. Airiti Library. https://doi.org/10.6833/CJCU.2014.00101
- 楊政樺、盧衍良、林詠晴、李靜旻（2013）。臺灣高鐵標準車廂及臺鐵自強號常乘顧客再購動機暨市場觀察：羅序模式之應用。顧客滿意學刊，9(1)，103-134。https://www.airitilibrary.com/Article/Detail?DocID=18164668-201303-201304110009-201304110009-103-134
- Wu, C. J. (2021). 基於深度學習之超高壓輸電線安全裕度預測及調度策略擬定 [master's thesis, National Taiwan University]. Airiti Library. https://doi.org/10.6342/NTU202103267
- 宋鴻康（2019）。Application of Trains Operation Strategy in Energy Saving Research of Taiwan Railway Rolling Stock〔博士論文，逢甲大學〕。華藝線上圖書館。https://doi.org/10.6341/fcu.P0100020
- 賴威宇（2011）。A Study on the Tapering Rate System for Taiwan Railway Administration〔碩士論文，國立交通大學〕。華藝線上圖書館。https://doi.org/10.6842/NCTU.2011.01114