Identifying the mechanisms by which aggressive behavior and road conditions affect congestion is essential for operational efficiency improvement. To address the frequent congestion in the exit area of expressways, this paper proposed an interpretable machine learning based approach to congestion causation analysis. The congestion index provided by the navigation software is used to identify the congestion status of the road. The XGBoost model is used to construct a congestion model, and the SHAP is used to analyze the impact of various influencing factors on the occurrence of congestion. The results show that the XGBoost can accurately identify the congestion of expressway exits with an precision of 82.5%, a recall of 82.1%, and an F1-score of 82%; the sharp acceleration and sharp deceleration behaviors are important influencing factors leading to congestion, and more sharp deceleration and sharp acceleration aggravate the induced occurrence of congestion.

Traffic congestion;Machine Learning;XGBoost;SHAP;Expressway exits;Aggressive driving behavior