As a core component of urban intelligent transportation infrastructure, the performance of transportation electromechanical system is directly related to the efficiency and safety of transportation operation. In this paper, an improved Canopy-K-means clustering algorithm is proposed to categorize traffic E&M systems, and a performance evaluation method is constructed based on the random forest model. The improved clustering algorithm adopts the “median and maximum distance product method” to determine the initial clustering center, and reduces redundant operations by optimizing the distance calculation. At the same time, a random forest evaluation model is established based on the driving performance index system to scientifically evaluate the performance of the electromechanical system. The experimental results show that the improved Canopy-K-means algorithm achieves an average accuracy of 83.48% on six UCI datasets, which is 5.85% higher than the traditional K-means algorithm; the running time is 169.53ms, which is 35.84% shorter than the traditional algorithm. The random forest model performs well in the evaluation, with an AUC value of 0.951 for the ROC curve and a KS value of 0.8044, which is significantly better than the traditional methods such as logistic regression. The SHAP analysis reveals that the features contributing the most to the evaluation are the absolute maximum of longitudinal acceleration, the mean value of longitudinal velocity, and the standard deviation of the angle of the heading angle from the centerline of the lane. This study provides an effective method for accurate classification and scientific assessment of transportation electromechanical systems.
