多元整合旅運服務(MaaS)使用者特徵分析：以高雄市為例

Abstract

本研究旨在探討多元整合旅運服務（Mobility as a Service, MaaS）之使用者公共運輸使用行為特性。MaaS 的核心理念係透過整合多種公共運輸系統、共享運具及副大眾運輸系統等，提供使用者無縫且便利的移動體驗，進一步提升都市居民對公共運輸與綠色交通系統的使用意願，藉以降低民眾對私人機動運具的依賴，促進城市的永續發展。透過理解 MaaS 之公共運輸使用行為可以理解公共運輸系統作為 MaaS 系統之服務核心，使用者的旅次特性得以作為後續服務改善的重要基礎。因此本研究以高雄市 MaaS 系統「 MeNGo」 之電子票證資料作為實證分析對象，透過對使用者公共運輸電子票證進行旅次特性之特徵工程、主成分分析（PCA）、使用者分群及多項羅吉特模型（MNL）等方法進行分析，欲了解在 MaaS 系統中之使用者公共運輸旅次特性。 本研究資料使用2023 年 11 月 MeNGo使用者之公共運輸電子票證資料，總共為 1,603,411 筆旅次交易紀錄，涵蓋 45,115 位使用者。研究中在資料中萃取出 32 項旅次特徵，並進一步歸納為 5 個主要行為構面與 8個使用者集群，依使用頻率區分為通勤族群（包含長距離規律、短距離規律、高轉乘、多目的使用者）與非通勤族群（混合模式、間歇性平日／假日、一次性使用者）兩大類別。分析結果發現，「使用頻率」與「時空規律性」為區分兩大族群之關鍵特性，其中通勤族群偏好班距穩定的捷運系統，於部份集群中呈現出高度的時空規律性；非通勤族群則有較多元且分散的運具選擇，並以公共運輸與私人運具混合使用的形式完成日常出行之目的。接續透過效果量分析，檢定各通勤族群間的使用行為差異程度，並透過多項羅吉特模型驗證站點特性與個人屬性對通勤族群使用行為的影響。研究之主要貢獻在於提出一套系統性的 MaaS 使用者電子票證資料處理與分析流程，並針對不同的使用者特性提出對應的改善策略，包括針對使用者行為特性進行套套票組合設計、提升公車服務接駁效率、以及推動跨機關資料整合與開放。

This study investigates the public transport trip characteristics of Mobility as a Service (MaaS) users based on Kaohsiung City’s MaaS platform MeNGo. MaaS aims to integrate various public transport systems, shared-mobility services, and para-transit modes to provide a seamless, convenient travel experience that enhances traveler’s accessibility to public transport, increases ridership intentions, reduces reliance on private motorized vehicles, and fosters sustainable urban development. Because public transport forms the core of MaaS ecosystem, understanding MaaS users’ trip characteristics and patterns can help of providing service improvement. Accordingly, this research employs smart-card data from Kaohsiung City’s MaaS platform “MeNGo” as a case study and applies trip-feature engineering, principal component analysis (PCA), K-Means++ clustering, and a multinomial logit model (MNL) to analyze public transport trip characteristics within a MaaS context. The dataset contains 1,603,411 public transport transactions made by 45,115 MeNGo users in November 2023. Thirty-two trip features were extracted and synthesized into five behavioral dimensions, from which eight user clusters have been identified. Based on usage frequency, these clusters were grouped into commuter segments (long-distance regular, short-distance regular, high-transfer, and multi-purpose users) and non-commuter segments (mixed-mode, intermittent weekday/holiday, and one-off users).Usage frequency and spatial-temporal regularity emerge as the decisive attributes separating the two macro-segments: commuter clusters favor the metro system’s stable headways and display pronounced regularity, while non-commuter clusters adopt more diverse and dispersed modal portfolios, often combining public and private modes to meet daily travel needs. Effect-size analysis was performed to gauge behavioral differences among commuter clusters, and an MNL was estimated to verify how station attributes and individual characteristics influence commuter behavior. The study’s primary contribution includes formulation of a comprehensive and systematic workflow for processing and analyzing MaaS smart-card data. Policy recommendations include designing ticket bundles tailored to various user behaviors and their needs, improving bus feeder efficiency, and promoting cross-agency data integration and open-data initiatives.