Sarah Kaleem (Prince Sultan University, PSU) Awais Ahmad (Imam Mohammad Ibn Saud Islamic University, IMSIU), Muhammad Babar (Prince Sultan University, PSU), Goutham Reddy Alavalapati (University of Illinois, Springfield)
This paper presents an integration of Federated Learning (FL) with Big Data Analytics (BDA) for Intelligent Transportation Systems (ITS). By leveraging the decentralized nature of FL, the framework enhances privacy, reduces latency, and improves scalability, addressing key limitations of traditional BDA approaches. This research demonstrates the potential of FL to revolutionize data analytics in ITS by enabling realtime applications and facilitating personalized insights. The key contributions of this research include the integration of FL with BDA to tackle traditional BDA challenges, the implementation of FL algorithms within the proposed integrated framework, and a comprehensive performance and scalability analysis. Additionally, the paper presents the development and validation of a specialized ITS dataset designed for FL environments. These contributions collectively highlight the transformative potential of FL in optimizing traffic management and public transportation systems through efficient and scalable data analytics. We demonstrate FL’s capability to efficiently manage and analyze ITS data while maintaining user privacy and scalability. Our findings reveal that FedProx achieved the highest global accuracy at 79.61%, surpassing FedSGD at 79.10% and FedAvg at 78.01%.