Seminar: Cluster-Centric Anomaly Detection and Characterization in Spatial Time Series

Dr. Hesam Izakian
University of Alberta

Cluster-Centric Anomaly Detection and Characterization in Spatial Time Series

Candidate for Faculty Position in Computer Science

Department of Computer Science
Monday, October 27, 2014, 2:30 p.m., Room EN-2022


 

Abstract

Anomaly detection in spatial time series is a challenging problem with numerous potential applications in environmental, engineering, economic, and health sciences. A comprehensive anomaly detection approach not only should be able to detect and identify emerging anomalies, but it has to characterize the essence of anomalies by visualizing the structures revealed within data.

In this talk, a cluster-centric framework for anomaly detection and characterization in spatial time series is introduced. For this purpose, the time series part of data is divided into a set of subsequences and the available spatio-temporal structures within the generated subsequences are discovered through a fuzzy clustering technique. Since in spatial time series, each datum is composed of features dealing with the spatial and the temporal (one or more time series) components, clustering of data of this nature poses some significant challenges, especially in terms of a suitable treatment of different components of the data. An extended version of the Fuzzy C-Means (FCM) clustering by introducing a composite distance function with adjustable weights (parameters) controlling the impact of different components in the clustering process is proposed, and a reconstruction criterion is used as a vehicle to quantify the performance of the clustering method.

By comparing the revealed structures (clusters) in spatial time series in successive time intervals, one assigns an anomaly score to each cluster measuring the level of unexpected changes in data. Moreover, through developing some fuzzy relational dependencies, the propagation of anomalies can be visualized in an understandable way to end-users. To illustrate the proposed technique, an outbreak scenario has been considered. Experimental studies show that the proposed technique is able to find incident anomalies and quantify the propagation of anomalies over time.

Contact

Department of Computer Science

230 Elizabeth Ave

St. John's, NL A1B 3X9 CANADA

Tel: (709) 864-2530

Fax: (709) 864-2552

becomestudent@mun.ca