FEDEVD: A FEDERATED ROAD TRAFFIC EVENT DETECTOR FROM SOCIAL NETWORKS

Authors

Ahmad Traboulsi, Department of Mathematics and Computer Science, Faculty of Science, Beirut Arab University, Debbyeh, LebanonFollow
May Itani, Department of Mathematics and Computer Science, Faculty of Science, Beirut Arab University, Debbyeh, LebanonFollow
Layal Abu Daher, Department of Mathematics and Computer Science, Faculty of Science, Beirut Arab University, Debbyeh, LebanonFollow
Ali Haidar, Electrical and Computer Engineering, Faculty of Engineering, Beirut Arab University, Debbyeh, LebanonFollow

Keywords

Federated Learning, Event Detection, Road Traffic, Online Social Networks

Disciplines

Engineering | Physical Sciences and Mathematics

Abstract

The increasing population and the corresponding rise in the number of vehicles, coupled with inadequate public transportation, is increasing the already existing traffic problem. One solution to this problem has been the implementation of Traffic Monitoring Systems (TMS). However, TMS deployment entails significant costs, including (1) the installation of hardware in key areas and (2) the employment of dedicated monitoring personnel. Simultaneously, Online Social Networks (OSNs) have become global platforms with rapidly growing user bases, leading to a surge in user engagement. This widespread participation has transformed social networks into invaluable sources of data for analytics, business intelligence, and decision support. In the past years, multiple contributions have been made regarding event detection systems from social networks covering different aspects of the problem, like diversity – of social networks, language and dialects-, accuracy and reliability of detection systems, visualization techniques, etc… In this paper, we propose a framework for traffic event detection from online social networks to (1) increase the detection rate, (2) optimize performance, and (3) improve overall security. Our proposed solution replaces conventional methods of managing event detection systems through decentralizing the detection process by leveraging federated learning techniques. We implemented a test bench that trains a Pytorch model in the FLower federated framework, the implemented solution was evaluated over a dataset obtained by web scraping the X social network website, providing a proof of concept to build distributed environments that constantly train event detection models and share them with collaborating trainers.

Author ORCID Identifier

Ahmad Traboulsi https://orcid.org/0009-0001-2696-2174

May Itani https://orcid.org/0000-0002-0738-0822

Layal Abu Daher https://orcid.org/0000-0002-4041-3243

Ali Haidar https://orcid.org/0000-0001-8065-3658

Recommended Citation

Traboulsi, Ahmad; Itani, May; Abu Daher, Layal; and Haidar, Ali (2025) "FEDEVD: A FEDERATED ROAD TRAFFIC EVENT DETECTOR FROM SOCIAL NETWORKS," BAU Journal - Science and Technology: Vol. 6: Iss. 2, Article 5.
DOI: https://doi.org/10.54729/2959-331X.1154

ISSN

2959-331X