比較k-NN模式與時變係數模式對高速公路旅行時間預測之研究

A Study of Comparison of k-NN Model and Time-Varying Coefficient Model for Predicting Travel Time on Freeways

陳建旻(Jian-Min Chen)

指導教授 ： 王晉元

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.2009.00593

摘要

近年來政府大力推動智慧型運輸系統(Intelligent Transportation System, ITS)，而先進用路人資訊系統(Advanced Traveler Information System, ATIS)即為ITS中的一子系統。在ATIS中，為了要給予用路人準確的資訊，以作為路徑、運具選擇之依據，路徑旅行時間的預測是一項重要的課題。尤其在高速公路路網建置完成後，適當交通預測資訊之提供對用路人行為決策更顯得其重要，不僅可以作為駕駛者選擇適當之路徑與出發時間憑藉之依據，用路者亦可藉此選擇最短之旅行時間到達目的地，以真正發揮高速公路路網之整體績效。
本研究係針對國內高速公路為研究對象，利用探針車所蒐集到的即時交通資料，分別利用k-NN模式與時變係數模式預測未來旅行時間，並針對兩種模式的預測結果進行績效評估並作比較，以期能提供精準之旅行時間預測，提供用路人路徑選擇或是出發時間決策判斷之依據。
本研究以國道一號高速公路楊梅到泰山收費站作為實際測試對象，經由測試結果可得知，本研究所構建的k-NN與時變係數兩種旅行時間預測模式，都是屬於高精準預測，而k-NN模式又能夠得到比時變係數模式更好的預測結果。因此，從測試結果顯示出本研究的預測方法可實際應用在高速公路上，且可提供用路人精準的旅行時間預測。

關鍵字

k-NN模式 ； 旅行時間預測 ； 時變係數模式 ； 探針車

並列摘要

In recent years, the government is actively promoting Intelligent Transportation System (ITS), and Advanced Traveler Information System (ATIS) is a subsystem of ITS. Travel time prediction is a very important of ATIS. When drivers have to make a decision, it is more important for drivers to use suitable traffic information. Traffic information will allow drivers to select appropriate routes and departure time to avoid congestion and arrive in the destination with the shortest time.
In this study, the probe vehicles collect real-time traffic information, and use the k-NN model and Time-Varying Coefficients (TVC) model to predict the future travel time, respectively. Evaluation and Comparison of two models for forecasting the results, hope to provide accurate forecasts of travel time to travelers departure time or route choice decision-making judgements based on.
We use the 1st National Freeway Yangmei to Taishan Toll Station as the actual test object. The testing results show that k-NN model and TVC model are high precision prediction, and k-NN model predict better than TVC model. The prediction method can actually use on the highway, and can provide accurate prediction of travel time to drivers.

並列關鍵字

k-NN model ； TVC model ； travel time prediction ； Probe Vehicle

參考文獻

[23]. 蔡百里，「資料融合技術應用於旅行時間推估之研究」，淡江大學運

[1]. Altman, N. S., “An Introduction to Kernel and Nearest Neighbor Nonparametric Regression,” The American Statistician, Vol. 46(3), pp.175-185, 1992.

[4]. Clark, S., “Traffic Prediction Using Multivariate Nonparametric Regression,” Journal of Transportation Engineering, Vol.129, No.2, pp.161–168, 2003.

[5]. Kwon, J., Coifman, B., and Bickel, P., "Day-to-Day Travel Time Trends and Travel Time Prediction from Loop Detector Data," Transportation Research Record no.1717, Transportation Research Board pp. 120-129, 2000.

[6]. Lam, W.H.K., Tang, Y.F., Chan, K.S., and Tam, M.L., “Short-term hourly traffic forecasts using Hong Kong annual traffic census,” Transp., Vol. 33, No. 3, pp. 291-310, 2006.