Theodoraki E. M., Koliou P., Yannis G., “Enhancing Risky Driving Behavior Classification Using Conditional GANs (cGANs): A Data Augmentation Approach”, Proceedings of the Transport Research Arena TRA 2026 Conference, Budapest, 18-21 May 2026.

Accurately classifying risky driving behaviors is essential for road safety, but is hindered by class imbalance, with dangerous behaviors underrepresented in most datasets. This study investigates the use of Conditional Generative Adversarial Networks (cGANs) to generate synthetic data for rare risk categories and improve classification performance. Driving behaviors were categorized as Normal, Dangerous, or Avoidable Accident. Following augmentation, a binary classification scheme (Normal vs. Avoidable Accident) was adopted. A cGAN was trained to generate synthetic high-risk scenarios, which were used to balance the dataset. Classifiers—GAN-based, XGBoost-RF, and RNN-AdaBoost—were evaluated on both original and augmented datasets. Results showed that cGAN-generated data improved model accuracy (Belgium: 76%→90%; UK: 79%→91%). However, hybrid models achieved near-perfect accuracy on augmented data, indicating overfitting. This study highlights both the promise and risks of cGAN-based augmentation, emphasizing the need for careful validation to ensure real-world applicability.