AI-based anomaly detection in multimodal urban mobility data streams

Abstract

The growing availability of real-time urban mobility data enables new approaches for monitoring complex transport systems, but integrating heterogeneous data sources remains a significant challenge. This paper investigates the use of AI-based anomaly detection for identifying irregular patterns in multimodal urban mobility data from the Norwegian transport ecosystem. The analysis uses publicly available data obtained through APIs provided by Entur for public transport, Oslo Bysykkel for micro-mobility, Avinor for aviation- related mobility data, and Statens vegvesen for road traffic data. The proposed workflow integrates and preprocesses data from multiple operational sources before applying unsupervised anomaly-detection methods to identify unusual mobility fluctuations and delay-related inconsistencies across transport modes. The results indicate that anomaly patterns are more frequently observed during peak operational periods and temporary transport-system disruptions. The study demonstrates that unsupervised anomaly-detection methods can support operational monitoring of integrated urban mobility systems in near real-time environments.