Geographical and Network Hotspot Detection, Prioritization, and Early Warning

• Activity Number: 287
• Type: Invited
• Date/Time: Wednesday, August 11, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Environmental and Ecological Statistics
• Author(s): Charles Taillie*+ and Ganapati P. Patil
• Companies: Pennsylvania State University and Pennsylvania State University
• Keywords: geosurveillance ; hotspot rating ; multicriteria prioritization ; partially ordered sets ; spatial scan statistic ; typology of space-time hotspots

Abstract

Hotspot means something unusual-an anomaly, aberration, outbreak, elevated cluster, etc. The responsible factors may be natural, accidental, or intentional. This presentation describes the upper level set (ULS) scan statistic for hotspot detection across geographic regions and across networks. The method is computationally efficient and can identify clusters of arbitrary shape including those that may not be adequately captured by the traditional circle-based scan statistic. The ULS statistic extends to the space-time domain where, because it allows for arbitrarily shaped space-time hotspots, it can characterize the temporal evolution of a spatial hotspot. This leads to the "typology of space-time hotspots." Changing patterns of urban poverty in different metropolitan areas of the U.S. are examples of such typologies. We also describe methods for multi-criteria prioritization of identified hotspots, employing the notion of linear extensions of partially ordered sets. These innovative tools are relevant to syndromic surveillance, cyber surveillance, crop surveillance, invasive species surveillance, drinking water system surveillance, network surveillance, among others.