Abstract
Collecting real-world mobility data is challenging. It is often fraught withprivacy concerns, logistical difficulties, and inherent biases. Moreover,accurately annotating anomalies in large-scale data is nearly impossible, as itdemands meticulous effort to distinguish subtle and complex patterns. Thesechallenges significantly impede progress in geospatial anomaly detectionresearch by restricting access to reliable data and complicating the rigorousevaluation, comparison, and benchmarking of methodologies. To address theselimitations, we introduce a synthetic mobility dataset, NUMOSIM, that providesa controlled, ethical, and diverse environment for benchmarking anomalydetection techniques. NUMOSIM simulates a wide array of realistic mobilityscenarios, encompassing both typical and anomalous behaviours, generatedthrough advanced deep learning models trained on real mobility data. Thisapproach allows NUMOSIM to accurately replicate the complexities of real-worldmovement patterns while strategically injecting anomalies to challenge andevaluate detection algorithms based on how effectively they capture theinterplay between demographic, geospatial, and temporal factors. Our goal is toadvance geospatial mobility analysis by offering a realistic benchmark forimproving anomaly detection and mobility modeling techniques. To support this,we provide open access to the NUMOSIM dataset, along with comprehensivedocumentation, evaluation metrics, and benchmark results.