2020/10/06 by Seyyid Emre Sofuoglu, Sofuoglu, Seyyid Emre, Selin Aviyente +1
Computer Science · Engineering · Mathematics · #Anomaly Detection Techniques and Applications #FOS: Electrical engineering #Signal Processing (eess.SP) #Tensor decomposition and applications #Traffic Prediction and Management Techniques #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2010.02889
openalex publication_date 2020/10/06 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Anomaly detection in spatiotemporal data is a challenging problem encountered\nin a variety of applications including hyperspectral imaging, video\nsurveillance and urban traffic monitoring. In the case of urban traffic data,\nanomalies refer to unusual events such as traffic congestion and unexpected\ncrowd gatherings. Detecting these anomalies is challenging due to the\ndependence of anomaly definition on time and space. In this paper, we introduce\nan unsupervised tensor-based anomaly detection method for spatiotemporal urban\ntraffic data. The proposed method assumes that the anomalies are sparse and\ntemporally continuous, i.e., anomalies appear as spatially contiguous groups\nof locations that show anomalous values consistently for a short duration of\ntime. Furthermore, a manifold embedding approach is adopted to preserve the\nlocal geometric structure of the data across each mode. The proposed framework,\nGraph Regularized Low-rank plus Temporally Smooth Sparse decomposition (GLOSS),\nis formulated as an optimization problem and solved using alternating method of\nmultipliers (ADMM). The resulting algorithm is shown to converge and be robust\nagainst missing data and noise. The proposed framework is evaluated on both\nsynthetic and real spatiotemporal urban traffic data and compared with baseline\nmethods.\n