A Framework to Predict Freeway Traffic Speed in Snowy Weather

Abstract

Providing accurate and reliable predictive traffic information is fundamental for the successful operation of proactive traffic management and information strategies. In particular, traffic prediction during snowy weather conditions is of keen interest because this information results in a greater capacity reduction and crash occurrence. This study proposes a methodology for predicting freeway traffic speed in snowy weather. One feature of the proposed methodology is the incorporation of historical and real-time traffic speed patterns in the prediction of freeway traffic speed. The proposed methodology consists of three components. The first and second components use traffic speed patterns that can be obtained from the analysis of historical data recorded in the database at traffic management centers (TMCs). The last component uses one-step-ahead predicted traffic speed based on real-time data. Both road weather information systems (RWIS) and vehicle detection systems (VDS) data obtained from the Seoul-Chooncheol Freeway in Korea were used in the development of the model and the performance evaluation of transferability. The proposed methodology showed a promising result. Under normal conditions, 3.00% mean absolute percentage error (MAPE) was obtained. In addition, MAPEs of 4.70 and 6.83 were achieved during slight and heavy snow conditions, respectively. The findings of this study can be used to provide predicted traffic speed information as a component of traffic management and information systems for users, operators, and decision makers.