Olusola Agbebiyi
- Singidunum University, Belgrade, Serbia
Abstract:
This paper presents an AI enhanced early warning framework for monitoring compliance under the European Unionβs Digital Markets Act (DMA). A fine tuned BERT model classifies consumer discussions into Article 6 violation types (self preferencing, bundling, discrimination, interoperability restriction) and combines results with community level statistical validation. On Reddit data (n = 82; 2020β2025), the classifier achieved 86.2% accuracy and macro F1 = 0.84; community validation averaged 0.86 with Z = 2.99 (p < 0.01). Projected operational costs decrease by ~90% relative to manual monitoring. The framework provides a reproducible, low cost pipeline for proactive DMA surveillance aligned with Articles 5, 6, 19, and 20 and supports prioritization of investigations through confidence weighted alerts.
International Scientific Multidisciplinary Conference: AI for a Smarter Tomorrow - AI-SMART , September 25-26, 2025
Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-Non-Commercial 4.0 License (https://creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work without further permission.