Anomaly Detection in Space-Time (and higher dimensional) Point Processes
Michael D. Porter, (NCSU), email@example.com
There is a growing need to develop methodologies for change detection in space-time processes. This talk discusses some approaches to anomaly detection (a specific type of change where the change occurs in a local region of space) in space-time point processes. The problem of detecting such changes is applicable in areas such as disease surveillance, computer intrusion detection, target detection, and crime and terrorism.
We take a likelihood based approach where the unknown pre and post change parameters are estimated adaptively, thus expanding the common GLR, CUMSUM, and Shiryaev-Roberts change detection methodologies. As one of the post-change parameters is the region where change has occurred, we also discuss some methods to identify this region in 2-D and higher dimensional spaces.