高速公路旅行時間預測-以k-NN法及分群方法探討

Freeway Travel Time Prediction by Using the k-NN Method and Comparison of Different Data Classification

蔡繼光(Chi-Kuang Tsai)

指導教授 ： 卓訓榮

國立交通大學/管理學院/運輸科技與管理學系/碩士(2009年)

https://doi.org/10.6842/NCTU.2009.00575

摘要

旅行時間預測為先進旅行者資訊系統的一部分，其目的為利用即時的速度、流量等交通資訊，即時地、準確地預測旅行時間以供用路人、交通管理人員進行查詢或各種決策分析。對於用路人而言，大多以旅行時間最小為其決策目標進行路線規劃。然而用路人只能掌握部分的路況資訊，較難決定旅次之出發時間、並對路徑選擇進行最有效率的決策與預估，降低運輸行為之不確定性。因此旅行時間預測為智慧型運輸系統中一個重要的課題。
本研究利用電子收費系統之歷史資訊及高速公路偵測器所收集到的即時與歷史交通資訊(速度、流量)進行分析。在研究方法上，利用比對的方法，找出與即時交通資訊類似之歷史交通資訊，再利用電子收費系統之資訊推估該歷史時間點之旅行時間，進行旅行時間預測。
本研究在資料比對的方法上採用k-NN法(k-Nearest Neighbor Method)進行處理。在資訊比對上。為了進一步提高上述方法的準確率，由單時間點的比對擴大為，比對一段時間的交通資訊的變化，並考量在不同偵測器下交通特徵的差異，因此再加入偵測器的權重以進行分析。並討論在不同的資料分群下，是否可以提升預測準確率。
最後利用台灣國道三號高速公路為對象進行研究，驗證本研究方法是否有足夠能力得到準確的預測結果。

關鍵字

k-NN法 ； 高速公路旅行時間預測 ； 資料分群

並列摘要

The travel time prediction is a part of Advanced Traffic Information Systems, ATIS. It is using the instant speed and flow information to predict the travel time on certain the path. It helps the road user and traffic manager to survey or strategy analysis. Using the travel time information, it is useful for road user to do the trip planning by minimizing the travel time. Without travel time information, it is hard to gather enough information doing the efficient trip planning, nor estimating the travel time. Therefore, travel time prediction is the important issue in intelligent transportation system. It help trip planning and decrease the uncertainty of the trip.
The study uses the historical electronic toll collection, ETC, information and historical and real-time vehicle detector, VD, information, speed and flow, on the freeway. Using the comparison method find out the familiar traffic information between historical data and real-time data. Then, estimate the travel time of the historical time by ETC data. With the familiar period and estimated travel time predict the real-time travel time.
The study adopts the k-nearest neighbor method, k-NN method, in data comparison. For decreasing the rate of error, instead of comparing each time period compares the traffic information trend to find out the similar data. Besides, we consider the difference between different VDs, and weight the VD data to adjust. Finally, discussing different data classification increases the performance of prediction.
Finally, the study took the third freeway in Taiwan for example. Figure out if the method got enough predicting ability.

並列關鍵字

k-NN method ； freeway travel time prediction ； data classification

參考文獻

[19]魏健宏，林士傑，李穎「高速公路客運車輛旅行時間預測之實證評析」，運輸計劃季刊，第32卷，第4期，651~679頁，2003。

[2] van Berkum, E., and P. van der Mede, “The Impact of Traffic Information: Dynamics in Route and Departure Time Choice, Delft University Press, Delft, the Netherlands”, PhD. thesis of the Transport and Planning Department, Faculty of Civil Engineering and Geosciences, Delft Unive rsity of Technology, 1993.

[4] Kwon, J., B. Coifman, and P. Bickel,“Day-to-Day Travel-Time Trends and Travel-Time Prediction from Loop-Detector Data” , Transportation Research Record: Journal of the Transportation Research Board, Vol.1717,pp.120-129,2000.

[5] Sen, A., P. Thakuriah, and X.Q. Zhu, “Frequency of Probe Reports and Variance of Travel Time Estimates”, Journal of Transportation Engineering, Vol. 123, No. 4, pp. 290-297, 1997.

[6]Rice, J. and E. van Zwet, “A Simple and Effective Method for Predicting Travel Times on Freeways”, IEEE Transactions on Intelligent Transportation Systems, Vol. 5(3), pp. 200-207, 2004.