運用悠遊卡巨量資料分析公車乘客行為

Analysis of Bus Passengers’ Travel Behavior based on Easycard Big Data

謝萬興(Wan-Hsing Hsieh)

指導教授 ： 張學孔

國立臺灣大學/工學院/土木工程學研究所/碩士(2015年)

https://doi.org/10.6342/NTU.2015.00283

摘要

近年來，巨量資料分析應用在交通領域對於政府、營運者、乘客之效益逐漸獲得重視，其中智慧卡資料具備可儲存大量個人旅行資料、由卡號連接到使用者以及比現存運輸資料來源可以獲得更長期連續旅次資料的特性，具備相當高的商業價值。悠遊卡自2000年發行至今，累積發卡量已突破5,000萬張，目前平均每日應用於各領域之總交易筆數合計超過550萬筆。本研究係以悠遊卡刷卡資料輔以站點刷卡人數統計，由2014年11月1日至11月30日共4,973,841筆交易資料中，針對南京東西路廊公車群組共20路線進行分析，了解公車使用者之行為及特性，並探討不同族群在時間面與空間面的變異情形。
時間面分析結果顯示平日比起假日具有較明顯的尖峰，學生族群在平日下午8點至10點會出現另一使用尖峰。敬老族群在平日與假日的使用規律性十分相似，與全部票種、一般及學生族群不同。此外，捷運松山線通車後造成南京路廊整體路線日運量在平日下降14.65%，假日下降8.97%，對於一般以及學生族群衝擊較大，而其餘票種在週日的運量反而有所提升。在空間面分析，研究結果呈現不同票種在上、下午尖峰的站點選擇所不同，以及通車前後各站點搭乘人數的變化。本研究另以關聯法則的Apriori演算法探勘資料欄位間關係，發現各路線間在通車前、通車後與不同票種間捷運轉乘比例的差異，由通車前的11.47%下滑到通車後9.74%。本研究選取2014年11月第二週之平常日資料進行通勤旅次分析，建立出對應卡號在第一筆與最後一筆交易紀錄的使用時間及路線選擇矩陣，研究結果指出上午8至9點與下午6至7點為使用頻率最高的組合，而選擇搭乘同一條路線的比例54.96%，兩者的時間間隔，可呈現出不同族群的搭乘特性與機動力。
本研究建立出一套從資料前置處理、時間面分析、空間面分析、資料關聯性分析與通勤旅次分析之流程，對於任何使用智慧卡搭乘之公車路線皆可應用，具可應用性及可移轉性，可供其他城市參考。主管機關在新政策之實施、新路線之引進以及接駁路線的協調等議題時，可檢視各族群的使用特性、路線載客績效及變化輔助決策的進行，業者亦可據以進行營運策略的調整，提升整體服務品質。

關鍵字

智慧卡 ； 公共運輸 ； 使用者行為 ； 時空變異性 ； 巨量資料

並列摘要

In recent years, Big Data has been applied to transportation field and obtained benefits for governments, operators and passengers. Smartcard data with characteristics of plenty of personal travel data connected to cardholders as well as long-term continuous journey information, it has been considered having high commercial value. Easycard has issued 50 million cards since 2000, while the daily transactions were more than 5 million in 2014. The study aims to understand the behavior and characteristics of bus passengers, to explore spatial-temporal variability among different user groups. The study selected 20 bus routes in Nanjing Corridor for case study in which data from nearly 5 million transaction records in November 2014 were collected and analyzed.
Temporal analysis indicated that there had two obvious peaks on weekdays compared to holidays while student group had a third peak from 8PM to 10PM on weekdays. For all card types, adult and student groups had less regularity between weekdays and holidays, while elderly one remained high. It is also shown opening MRT Songshan Line has caused bus passengers in Nanjing corridor fell 14.65 % on weekdays and 8.97 % on holidays, especially for adult and student card types, while the rest card types (Concessionaire, Senior, Charity and Escort) yet increased on Sundays after MRT operation. The results of spatial analysis showed stop selection varies from different card types on morning and afternoon peaks. On associate rule analysis side, Apriori algorithm was applied to conduct data mining on relationships among data fields. It is found that MRT-Bus interchange proportion changed before and after operation of MRT Songshan Line among different card types and corridor bus routes, the entire routes fell from 11.47% to 9.47%.
The study selected over 1 million transaction records from November 10 to 14, 2014 to establish commuting time and route selection matrix which corresponded to the first and the last transaction records by matching unique card ID and sequence number of each card. The results showed that 8-9AM and 6-7PM was the highest frequency combination of all and there were 54.96% passengers choosing the same route on their first and the last transaction records. The duration between the two records could also showed the transit characteristics and mobility patterns among different card type users. Overall, the travel behaviour of the all categories follows a certain degree of temporal and spatial universality but also displays unique patterns within their own specialties.
The study proposed a systematic process from data pre-processing, spatial-temporal pattern analysis, association rule analysis to commuter journeys analysis. It has shown that the proposed methodology has high applicability and transferability for any bus routes with smart card as payment media. The process can help of evaluating impact of the implementation of the new policy, the introduction of new routes and coordination of feeder bus routes. It can also provide information of travel characteristics of each focused groups and changes on operating performance so that transport authorities on decision-making, operator efficiency and service quality could be all enhanced.

並列關鍵字

Smartcard ； Public Transport ； Travel Behavior ； Spatial-Temporal Heterogeneity ； Big Data

參考文獻

52. 羅惟元，「以悠遊卡交易資料探索公車路線之旅客起迄」，淡江大學運輸管理研究所碩士論文，2008年。

51. 廖宜靖，「台鐵運用智慧卡票證系統之策略與效益評估」，國立臺灣大學土木工程學研究所碩士論文，2008年。

49. 張恭碩，「非接觸式智慧卡於停車收費之效益分析」，國立臺灣大學土木工程學研究所碩士論文，2007年。

42. 吳沛樵，「非接觸式智慧卡對政府效益之分析」，國立臺灣大學土木工程學研究所碩士論文，2007年。

41. 朱正忠，「悠遊卡真的方便嗎?」，國立中央大學資訊管理學系碩士論文，2011年。