Expressway Traffic Speed Prediction Method Based on KF-GRU Model via ETC Data

Accurate prediction of expressway traffic speed can not only improve traffic efficiency and reduce road congestion, bring convenience to travelers, but also help management departments plan and manage road resources more effectively. Through the utilization and extension of expressway ETC gantry transaction data, this paper proposes an expressway traffic speed prediction method combining Kalman Filter (KF) and Gated Recurrent Unit (GRU) models. Build a vehicle trajectory dataset using ETC transaction data to generate section speed data set. Then, the KF is used to reduce noise and smooth the data, which solves the non-stationary and nonlinear problems of section speed. Then, the GRU is used to mine the traffic speed characteristics for traffic speed prediction, and finally, the experiment was validated using real ETC transaction data from Fuzhou to Xiamen. The results show that the KF-GRU method surpasses traditional models at different time intervals, thus verifying the correctness and reliability of the method.