PCIAFL: Personalized and Class Imbalance-Aware Federated Learning for Driver Behavior Classification
Abstract
Automated understanding of driver behavior from vehicular kinematics is vital for safety-aware intelligent transportation systems. However, centralized cloud processing suffers from latency, scalability, and privacy issues. Federated Learning (FL) provides a decentralized alternative but faces two major challenges: (i) non-IID client data due to heterogeneous driving styles and sensors, and (ii) severe class imbalance, as risky behaviors are inherently rare. In this work, we propose a personalized FL framework that uses a shared CNN-LSTM backbone with client-adaptive classifiers and incorporates a cost-sensitive loss to address behavior skew. Evaluated on the UAH-DriveSet dataset, our method achieves 92.60% accuracy and 91.68% macro-F1, outperforming FL baselines.
Recommended Citation
O. Osho et al., "PCIAFL: Personalized and Class Imbalance-Aware Federated Learning for Driver Behavior Classification," Icdcn 2026 Proceedings of the International Conference on Distributed Computing and Networking 2026, pp. 128 - 132, Association for Computing Machinery, Jan 2026.
The definitive version is available at https://doi.org/10.1145/3772290.3772312
Department(s)
Computer Science
Publication Status
Free Access
Keywords and Phrases
Class imbalance; Driving behavior; Non-IID data; Personalized FL
Document Type
Article - Conference proceedings
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 2026 Association for Computing Machinery, All rights reserved.
Publication Date
05 Jan 2026
Comments
Science and Engineering Research Board, Grant CRG/2022/005468