G-FedProx: Multi-Factor Traffic Flow Prediction Model based on Federated Learning

DOI

10.2991/978-94-6463-823-3_101How to use a DOI?

Keywords

Federal Learning; Traffic Flow Forecast; G-FedProx; Data Heterogeneity; LSTM

Abstract

In order to reduce the performance of the federated learning model caused by data heterogeneity in traffic flow prediction, this paper proposes an improved G-FedProx framework, and combines the real-time traffic status, weather and holiday data obtained by Amap Application Programming Interface (API) to build a multi-factor fusion Long-Short Term Memory (LSTM) prediction model. The traditional Federal average (FedAvg) algorithm is easy to cause global model deviation when the data is unevenly distributed, but G-FedProx introduces near-end term constraint to optimize the aggregation process, reducing the impact of client data differences on the model, and using dynamic communication compression to reduce transmission overhead. The experimental results show that, compared with FedAvg, the mean absolute error (MAE) of G-FedProx is reduced by 28.6% (2.89 ± 0.15 vs. 4.05 ± 0.32 km/h), and the communication cost is reduced by 20.1%. It also performs better in complex scenarios such as morning peak (error reduction of 29.6%) and rainy day (STCI index improvement of 22.4%). In addition, the model integrates meteorological and holiday characteristics, which significantly improves the spatio-temporal adaptability of the forecast. This study provides an efficient solution for distributed traffic prediction in smart cities, which can be expanded to multi-modal data fusion and edge computing optimization direction in the future.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yiming Li
PY  - 2025
DA  - 2025/08/31
TI  - G-FedProx: Multi-Factor Traffic Flow Prediction Model based on Federated Learning
BT  - Proceedings of the 2025 3rd International Conference on Image, Algorithms, and Artificial Intelligence (ICIAAI 2025)
PB  - Atlantis Press
SP  - 1045
EP  - 1057
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6463-823-3_101
DO  - 10.2991/978-94-6463-823-3_101
ID  - Li2025
ER  -