應用數位式行車紀錄器之時空資料建構旅行時間預測模式－以高速公路為例

A Travel Time Prediction Model Based on Tachograph's Time-Space Data-A Case of Freeway Syrstem

朱松偉(Song-Wei Chu) ； 鄭旭峰(Hsu-Feng Cheng) ； 吳耿毓(Ken-Yu Wu)

《健行學報》 34卷4期 (2014/10) Pp. 129-148

摘要

本研究為彙整產學計畫與專利技術移轉之產出成果的技術論文，係利用數位式行車紀錄器之資料進行旅行時間預測模式之建構。採用其資料分析並預測高速公路之旅行時間，以達到以下兩項目的1.利用數位式行車紀錄器之大量歷史資料，建構一套資料處理流程及2.推估不同星期從圓山交流道到桃園交流道的旅行時間。本研究所使用之歷史資料共高達5,123,622筆，因此必須研發一套適用於彙整與分析此龐大資料之處理流程，進而建構旅行時間之預測模式，同時亦透過統計SPSS軟體進行複迴歸分析與相關檢定。研究結果顯示平日各時段之平均旅行時間，迴歸模式與實際驗證之資料，其趨勢與分布相當一致，僅在時段五跟時段六出現比較大的誤差，最大誤差也不超過一分鐘，故可知本研究之迴歸模式具有相當高的準確性與實用性，同時對於高公局訂定交通管理策略與發布用路人資訊，以及客運公司之車隊管理規劃與商車管理均有莫大的助益。

關鍵字

數位式行車紀錄器 ； 旅行時間預測 ； 複迴歸分析 ； 車隊管理 ； 商車管理

並列摘要

This paper displays the output of projects and technical transfer. We use the data from Tachograph to predict travel time. In this study, the author adopted the data from Tachograph to probe into the travel time prediction of freeway to accomplish the purpose of the study. The objective of this study is to use the extensive recording data from Tachograph to develop a set of data processing. Second, is to predict the spending of travel time from Yuanshan interchange of National Highway No. 1 to Taoyuan interchange. There are5,123,622 historical data of this study. The study developed a set of data processing which is suitable the research to deal with theses historical data. The study used a great deal of historical data construction model which is the Regression model got from Multiple Regression Analysis from Statistics Package Social Sciences software. The outcome of the study show that the fluctuation of Regression Model accords with practical data in usual day, nothing but have more deviation in Interval Five and Interval Six. Although have more deviation in Interval Five and Interval Six, the maximum deviation is no more than one minute. So we can say that the Regression Model of the study have good accuracy and application. There are helpful for Management of freeway, Platoon Management of Bus Company and CVO.

並列關鍵字

tachograph ； travel time predict ； multiple regression analysis ； platoon management ； CVO