利用公車動態資料推估公車到站時間之研究

Abstract

先進公共運輸系統(Advanced Public Transportation System, APTS)中的公共運輸到站時間之推估，對於使用者來說可能是最具吸引力的資訊。準確的到站時間推估能減少使用者候車之時間和降低候車時的不安，藉此提升民眾的使用意願。因此到站時間推估對於公共運輸的發展扮演了重要的角色。本研究以公車作為研究對象，利用市區公車GPS的資料，建構一套能準確推估公車到站時間之模式。 本研究以k-NN方法為基礎，並針對歷史資料庫的插捕、距離量度比對時的門檻值設定、距離量度的調整、延遲加總進行探討。將測試結果分為6大類情境：整體、市郊區、平假日、尖離峰、平日尖離峰、距離站數，總共14種情境分別討論，並與未插捕之推估誤差進行差異性比較。由結果顯示本研究之推估模式於非尖峰時段有良好的表現，代表本研究之推估模式較適用在非尖峰時段。

The bus arrival time is the most needed information for public transit users. Providing bus arrival time can reduce the waiting time and the anxiety while waiting for a bus; it also improves the willingness to adopt the public transportation. Therefore, arrival time estimation plays an important role for the development of public transit systems. This research uses the data of urban buses to build a model, which estimates bus arrival time accurately. We proposed a k-NN based model for this purpose. Our k-NN model differs from the other similar models on the interpolation of historical database, distance threshold setting, and the distance adjustment. We designed 14 scenarios for testing purpose. The testing results showed that the proposed model has good performances on non-peak hours.