Prediction of Arrival Time for High-Frequency Vehicles at Expressway Toll Stations

Accurate prediction of vehicle arrival time at expressway toll station entrances remains a challenging issue in traffic flow modeling and management. To address this challenge, this work proposes an arrival time prediction model tailored for high-frequency vehicles. The model fully leverages expressway toll data, aggregates multi-dimensional features as inputs, and constructs a deep ensemble regression framework optimized by the Sparrow Search Algorithm (SSA). Specifically, SSA first optimizes a custom estimator to achieve adaptive parameter tuning. The optimized estimator is then fused with a deep forest model to significantly enhance predictive performance. Empirical studies conducted using real-world data from expressway ETC gantry systems demonstrate the proposed model's outstanding predictive accuracy, achieving a coefficient of determination (R²) of 0.918 and a low mean absolute error (MAE) of 0.027. These results validate the method's capability for high-precision arrival time prediction at the individual vehicle level for toll station entrances. This provides robust support for traffic operation status assessment, toll station scheduling optimization, and refined management of intelligent transportation systems.