Journal of the Eastern Asia Society for Transportation Studies, Vol.11, 2015 Application of Traffic State Prediction Methods to Urban Expressway Network in the City of Seoul Youngho Kim a, Woojin Kang b, Minju Park c a,b,c The Korea Transport Institute, 370 Sicheong-daero, Sejong-si, 339-007, Korea a E-mail:
[email protected] b E-mail:
[email protected] c Corresponding author: E-mail:
[email protected] Abstract: This paper proposes a traffic state prediction method based on two perspectives; short-term and long-term prediction. Modified KNN method is used for short-term prediction from recent 2 years of historical data set. Pattern of the day of the week is used to predict long-term. To overcome the gap between the result of short-term and long-term prediction, the weighted average for two predicted results is considered as the final predicted result. The proposed method is tested in the real urban expressway network and the performance of the proposed method is evaluated in this paper. Keywords: K-nearest neighbor, K-NN, short-term prediction, long-term prediction 1. INTRODUCTION Drivers usually precede a trip by checking out the traffic conditions for their route using computers, smart phones, or navigation systems. Even after they determine their travel route based on the information prior to the departure, they constantly look for the optimal travel route through navigation systems while driving. This is why the predicted traffic information is getting more important the advanced route planning. The predicted traffic information is produced in two steps, traffic state estimation and traffic state prediction.