SPATIO-TEMPORAL EVENT DETECTION BY INTEGRATING TOPOLOGICAL DATA ANALYSIS, INFORMATION DYNAMICS, AND SYSTEM STABILITY

Mohammad Hani, Pak Vadim G.

This paper introduces a new framework that combines dynamical systems theory, topological data analysis (TDA), and information theory to detect and analyze events in complex spatio-temporal datasets. By combining these metrics, the framework provides a robust, interpretable, and scalable solution for event detection. We demonstrate its application on the Milan 2013 dataset, which includes cellular activity, internet usage, and geolocalized Twitter data, demonstrating its ability to detect events such as social gatherings. Key innovations include: A unified approach combining robustness, topology, information theory, and Mahalanobis distance.

Keywords:Spatiotemporal, Loss of stability, Topological destruction, Information reconfiguration, Mahalanobis distance