The selection behavior of the lane-changing distance in the lane-changing process was examined by taking the lane-changing behavior in the urban expressway interchange weaving section as an object. Utilizing measured vehicle trajectory data, causal inference theory was applied to identify key factors influencing the choice of lane-changing distance. The factors included the velocity and lane-changing duration of the target vehicle before and after lane-changing, as well as the distance between the current lane and the vehicles in front and behind the target lane. The support vector model and the deep learning model were employed respectively to establish models for lane-changing distance selection behavior to test the validity of the influencing factor analysis. Results show that the screened factors contribute to enhancing the prediction speed of the behavioral choice model and improving the prediction accuracy of the deep learning model. While the support vector model offers faster predictions, it falls behind in prediction accuracy compared to the deep learning model. Conducting characteristic analysis on typical lane-changing behaviors provided a foundational basis for formulating management strategies in the urban expressway interchange weaving section, effectively supplementing existing research on lane-changing behavioral characteristics and offering a nuanced portrayal of the lane-changing process.