隨著都市結構與人口擴張，臺灣使用私有運具比例高達70.6%、大眾運輸使用僅佔18.1%，臺南市的大眾運輸使用率更只有6.7%，數據表示出臺南市大眾運輸使用上仍有進步空間。Hägerstrand的時間地理學指出，人類在日常生活中受限於各類物質及社會環境的制約，只能在有限的時間範圍內發生有限的行為；而瞭解人類在時空上移動的模式是為解決大眾運輸使用率低迷的首要步驟，才可將資源有效地安排，避免資源配置不當，造成資源浪費等狀況發生，也會使民眾對於大眾運輸的信心流失。過去傳統調查人類移動的方法中，常面臨到樣本數少且不夠精確的問題，無法全面瞭解城市中人類移動模式。本研究使用兩個月的票證紀錄剖析臺南市區公車路網，透過智慧卡所蒐集的大量旅次資料，能夠大幅提升資料數量及準確度，透過集群分析分類出乘客的搭乘模式，再使用關聯式法則分析路線區間的時空群聚及整體路網的使用狀況，以更細緻化的方式瞭解大眾運輸在臺南市的使用狀況。研究結果發現約75%的乘客的使用距離為5公里內；其餘乘客大多是以連結市區外圍的聚落為主，特別是在早上與下午的通勤時間；部分路線被識別出數量少但規律強烈的旅次，乘客可能受限於某些制約，例如：弱勢族群、學生族群等，應謹慎考量路線的存廢；反而是延駛或繞駛的路線未被識別出規律的旅次。因此透過這樣細緻的剖析，得出的結論與乘客的身分、公車路線上侷限程度、班次時刻表的安排有關，皆會影響到整體臺南市區公車的使用行為。經過全面性的瞭解使用分布狀況後，可提供給營運管理單位作為調整公車服務的依據，以鞏固既有路線所培養的客源為優先，並針對有問題的路線檢討改善，才能夠發揮大眾運輸最大效用。
With the expansion of urban structure and population, the proportion of private vehicle used in Taiwan is up to 70.6%, and leaves only 18.1% on public transportation system. In Tainan, the usage on public transportation is only 6.7%, which shows plenty of room for improvement. This sluggish utilization rate caused by a lack of confidence in the system results from the improper allocation of resources such as irrational distribution of bus stops or routes. To understand the patterns of human mobility in the city, the traditional methods investigated passenger behaviors but often faced the problem of small and inaccurate sample size. Using a two-month smart card dataset to analyze the bus network in Tainan City, this study explores the spatio-temporal patterns of passenger behaviors based on the time-geography approach and its implementation in GIS. The space-time paths, in this study, the O-D trips, are grouped into different clusters using K-means clustering in order to identify space-time characteristics of passenger behaviors. Moreover, for each cluster, an association rule mining is conducted to reveal more detail relations between the origins and destinations.
Our results show that about 75% of passengers traveled within 5 km, while the rest were long-distance commuters from the periphery of the city, especially during the morning and afternoon commuting time. Few crucial routes near the social welfare facilities are identified with high conviction rates which should not be axed for serving the disadvantaged groups, while some extended routes show no evidence of strong space-time usages from the result of association rule mining. Our discovery can be provided to the operator as a reference for future bus routes adjusting.
