Research on highway traffic flow prediction based on parallel time fusion transformation model

In order to meet the needs of refined traffic flow prediction in the environment of intelligent transportation system, a short-term traffic flow prediction model based on Parallel Temporal Fusion Transformer (PTFT) is proposed. The PTFT model adds the use of static variables, and effectively reduces the complexity of the model through the gating mechanism and variable selection network. The spatial correlation of traffic flow is considered, and the corresponding method is proposed for the interpretability of the model, and the basic framework of the model is constructed. In order to verify the prediction accuracy and reliability of the model, the traffic flow data collected by three traffic detectors on an expressway were selected for algorithm verification. The results show that the PTFT model proposed in this paper is superior to the commonly used models in predicting traffic flow. Incorporating weather variables is crucial for achieving higher prediction accuracy. Mastering the law of traffic flow under meteorological conditions is of reference significance for traffic guidance in severe weather. In addition, the model input considering the influence of periodic changes is crucial to the accuracy of the model prediction results.