Anomaly Detection and Characterization in Spatial Time Series Data: A Cluster-Centric Approach

Anomaly detection in spatial time series (spatiotemporal data) is a challenging problem with numerous potential applications. A comprehensive anomaly detection approach not only should be able to detect and identify the emerging anomalies but has to characterize the essence of these anomalies by visualizing the structures revealed within data in a way that is understandable to the end-user as well. In this paper, we consider fuzzy c-means (FCM) as a conceptual and algorithmic setting to deal with the problem of anomaly detection. Using a sliding window, the time series are divided into a number of subsequences, and the available spatiotemporal structure within each time window is discovered using the FCM method. In the sequel, an anomaly score is assigned to each cluster, and using a fuzzy relation formed between revealed structures, a propagation of anomalies occurring in consecutive time intervals is visualized. To illustrate the proposed method, several datasets (synthetic data, a simulated disease outbreak scenario, and Alberta temperature data) have been investigated.