臺北年輕機車騎士路徑選擇偏好：多重離散選擇模型之比較

Abstract

路徑選擇研究有助於瞭解道路使用者的偏好，並為國家運輸規劃政策提供參考，然而過去路徑選擇研究當中在方法與學理上仍存在空缺。方法上，方案相依性與偏好異質性為路徑選擇模式重要的兩個議題，亦各自發展出相應的解決方法，但同時處理兩者的方法卻尚未有定論。學理上，過去路徑選擇研究多以自行車與汽車為研究對象，且較侷限於自有運具，忽略對機車及共享運具的研究。機車因使用成本低，且具有穿梭小巷與車陣的機動性，成為臺灣與東南亞國家重要的運輸工具，而近年來共享運具的興起，其環保與經濟效益有益於提高城市居民的生活品質，兩者皆具有其研究的重要性。 本研究以臺北都會區20至30歲自有機車與共享機車騎士為研究對象，透過問卷蒐集其個人變數，並利用第三方手機應用程式記錄其騎乘過程之全球定位系統點位資料，利用ArcGIS 路網分析之功能建立使用路徑，並產生非時變最短路徑為替選路徑，同時使用Selenium套件與Google maps產生時間相依最短路徑作為另一替選路徑。研究方法結合兩種路徑規模因子與不同個體選擇模式進行模式擬合度比較，擇優進行後續路徑選擇分析，並探討自有機車及共享機車騎士路徑選擇偏好之異同。 本研究結果在方法上顯示路徑規模校正因子在處理方案相依性方面具有效性，且路徑規模校正羅吉特模式相較路徑規模羅吉特模式明顯提升模式擬合度。呼應先前研究，潛在類別模式在處理偏好異質性上優於混合羅吉特模式，然而與過去發現不同的是，在路線選擇問題上混合羅吉特模式並不明顯優於多項羅吉特模式。於學理上，本研究結果顯示機車騎士對於距離、右轉和混合住宅土地使用的偏好，可能與機車的高機動性相關。而在地騎士較偏好狹窄道路與右轉也顯示出地方知識在路徑選擇當中的作用。此外，共享與自有機車騎士的路徑選擇偏好具有差異，主要原因是前者具有較低的風險感知；然而，不同騎士群體間的差異仍然值得注意。以上發現改善現有路徑選擇研究在同時處理方案相依性與偏好異質性的方法效能，並填補對於機車騎士路徑選擇行為的知識缺口。

Route choice studies contribute to understanding road user preferences and provide a helpful reference for national transportation planning policies. However, research gaps exist in the methodology and theory of route choice research. Methodologically, solutions addressing alternative interdependence and preference heterogeneity, which are two crucial issues in route choice modeling, have been suggested individually in literature. Yet, consensus on simultaneously addressing both challenges remain elusive. From a theoretical perspective, prior studies predominantly focused on bicycles and cars as research subjects, primarily examining individually owned vehicles while overlooking scooters and shared transport systems. Scooters, given their low cost and maneuverability through narrow alleys and traffic, have become vital in Taiwan and Southeast Asian countries. Moreover, the recent surge in shared transport services provides environmental and economic benefits, substantially improving urban life quality. Consequently, investigating scooters and shared vehicles holds importance in research. This study focuses on owned and shared scooter riders aged 20 to 30 in the Taipei Metropolitan Area (TMA). We collected personal attributes of scooter riders through surveys and recorded GPS coordinates of their riding routes using a third-party mobile app. We estimated the riders’ actual routes and the time-invariant shortest alternative routes through ArcGIS network analysis. In addition, Selenium and Google maps were used to generate time-dependent shortest routes as other alternatives. We compared the goodness-of-fit of models combining two path size factors with different discrete choice models and selected the best-fit model for subsequent analysis. The similarities and dissimilarities in route choice preferences between riders of owned scooters and shared scooters were also explored. This study emphasizes the effectiveness of path size correction factors in addressing interdependence and confirms substantial improvements in the Path Sized Correction Logit model goodness-of-fit compared with the Path Size Logit model. The study reinforces the superiority of Latent Class Model over Mixed Logit Model in capturing heterogeneities and suggests that the Mixed Logit Model may not be remarkably better than the Multinomial Logit model in route choice studies. Theoretically, observations reveal the scooter riders’ preferences for distance, right turns, and mixed residential land use, attributed to the high mobility of scooters. The study underscores the importance of local knowledge, particularly among native riders favoring narrow roadways and right turns. In addition, distinctions in route preferences emerge between shared and owned scooter riders, highlighting the former’s lower risk sensitivity. However, the differences between rider segments still require attention. Overall, these findings offer valuable insights into the route choice behavior of scooter riders, thereby enriching the existing literature on route choice modeling and enhancing our understanding of scooter riders’ preferences.