Research on machine learning-based road traffic condition classification and prediction

To enhance the accuracy and real-time capability of road traffic status recognition and trend prediction, an ensemble machine learning model integrating XGBoost, LightGBM, and SVM algorithms was constructed to establish a multi-model discrimination system. This study investigates discrimination and prediction methods based on feature selection and principal component dimension reduction. Results demonstrate that the ensemble model outperforms individual models across Accuracy, Macro-F1, and Kappa metrics, exhibiting robust category discrimination and stability in the PCA projection space. This validates the model's generalization capability and application potential.