![[2407.12226] Individualized Federated Learning for Traffic Prediction with Error Driven Aggregation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2407.12226] Individualized Federated Learning for Traffic Prediction with Error Driven Aggregation
Summary
The paper presents NeighborFL, an individualized federated learning approach for traffic prediction that enhances real-time model updates and accuracy by addressing non-IID data characteristics.
Why It Matters
As urban traffic management becomes increasingly reliant on accurate predictions, this research offers a significant advancement in federated learning methodologies. By improving real-time adaptability and accuracy, NeighborFL can enhance smart city infrastructure and traffic flow management, ultimately contributing to more efficient urban mobility solutions.
Key Takeaways
- NeighborFL introduces individualized models for traffic prediction, enhancing accuracy.
- The approach utilizes error-driven aggregation to tailor predictions to local traffic conditions.
- Simulations show a 16.9% reduction in MSE compared to traditional federated learning methods.
- Real-time updates enable continuous adaptation to changing traffic patterns.
- The method addresses the challenges of non-IID data in traffic monitoring.
Computer Science > Machine Learning arXiv:2407.12226 (cs) [Submitted on 17 Jul 2024 (v1), last revised 16 Feb 2026 (this version, v2)] Title:Individualized Federated Learning for Traffic Prediction with Error Driven Aggregation Authors:Hang Chen, Collin Meese, Mark Nejad, Chien-Chung Shen View a PDF of the paper titled Individualized Federated Learning for Traffic Prediction with Error Driven Aggregation, by Hang Chen and 3 other authors View PDF Abstract:Low-latency traffic prediction is vital for smart city traffic management. Federated Learning has emerged as a promising technique for Traffic Prediction (FLTP), offering several advantages such as privacy preservation, reduced communication overhead, improved prediction accuracy, and enhanced adaptability to changing traffic conditions. However, majority of the current FLTP frameworks lack a real-time model updating scheme, which hinders their ability to continuously incorporate new incoming traffic data and adapt effectively to the changing dynamics of traffic trends. Another concern with the existing FLTP frameworks is their reliance on the conventional FL model aggregation method, which involves assigning an identical model (i.e., the global model) to all traffic monitoring devices to predict their individual local traffic trends, thereby neglecting the non-IID characteristics of traffic data collected in different locations. Building upon these findings and harnessing insights from reinforcement learning, we propose ...