了解旅運者其運具使用行為與旅次起迄資訊常是交通管理者十分重視的議題,於交通政策制定與營運管理階段,皆須先了解旅運者運具使用行為後,方能做後續交通分析與評估。傳統的運具使用行為研究多採用實地車流量調查、問卷發放與電話訪問等抽樣調查方式進行,往往耗費過多的人力與成本。本研究以手機信令資料為基礎,研擬一套推估旅運者全天移動軌跡中各時段運具使用行為之運具判斷方法。 本研究依照公車、捷運與台鐵之運具特性,結合旅運者信令資料與公共運具動靜態資料之時空軌跡,分別建立三種運具之運具判斷模式,最後擷取旅運者全天各旅次之運具別。於實測分析階段,透過實際信令資料的收集與指標設計,進行模式參數調整與驗證。研究結果顯示,此模式平均運具判斷率達88.1%、平均時間與空間判讀率分別為91.5%與86.8%,說明此模式可正確判斷出旅運者搭乘之運具別與旅次乘車時間。
Understanding the mode choice of one traveler has been a critical issue among transportation-related research. However, the data in previous research was collected by costly methods including traffic flow investigation, questionnaire survey, or telephone interviewing. Hence, a new source of data, cellular data, is used in this study, becoming the input data of mode detection models. The mode detection models are constructed individually by modes since the characteristics between bus, metro, and train are distinctive. By comparing the trajectory of a traveler’s cellular data and the real-time data of transits provided by PTX, the models could discriminate the modes of different trips. Also, this study proposes an investigation and defines indicators for validating the correctness of the models. The result shows that in average the accuracy of detection is 88.1% and that the mean ratios of time and space completion of trips are 91.5% and 81.6%, respectively. This interprets that the models not only can detect the mode of the traveler correctly but also can obtain the integrity of the trip accurately.