Abstract :

Predicting Driver Traffic Violations Is Essential For Enhancing Road Safety And Advancing Intelligent Transportation Systems. This Paper Proposes A Hierarchical Graph Neural Network (HGNN) Framework To Model Driver Behavior And Forecast Potential Violations. Real-world Traffic Datasets Are Preprocessed Into Multidimensional Indicators That Capture Behavioral, Contextual, And Temporal Patterns. At The Lower Level, Convolutional Neural Networks (CNNs) Extract Short-term Patterns, While Long Short-Term Memory (LSTM) Networks Capture Sequential Dependencies. These Features Are Then Integrated Into A Hierarchical Graph Attention Mechanism To Learn Spatial–temporal Interactions Between Drivers And Violation Types. A Self-adaptive Calibration Of Indicator Weights Further Improves Prediction Accuracy Across Diverse Traffic Contexts. Experimental Results Show That The HGNN Framework Achieves Superior Performance Compared To Conventional Deep Learning And Non-hierarchical Methods, Demonstrating Its Effectiveness In Building Safer Connected Vehicle And Smart Transportation Environments. Index Terms: Hierarchical Graph Neural Networks (HGNN), Traffic Violation Prediction, Driver Behavior Analysis, Spatial–Temporal Modeling, Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM), Graph Attention Networks, Intelligent Transportation Systems, Connected Vehicles, Road Safety

Published:

07-10-2025

Issue:

Vol. 25 No. 10 (2025)

Page Nos:

9-16

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite