|
Activity Number:
|
431
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
IMS
|
| Abstract - #305539 |
|
Title:
|
CitySense: Multiscale Space Time Clustering of GPS Points and Trajectories
|
|
Author(s):
|
Markus Loecher*+ and Tony Jebara and David Rosenberg
|
|
Companies:
|
Sense Networks and Columbia University and Sense Networks
|
|
Address:
|
110 Greene Street, New York, NY, 10012,
|
|
Keywords:
|
scan statistic ; geosurveillance ; spatial cluster detection ; overdispersed count data ; GPS ; CitySense
|
|
Abstract:
|
We propose a generalized scan statistic which performs periodic geographic surveillance of various measures that are relevant to cities. Our algorithm scans for elliptical and rectangular clusters. We optionally adjust for covariates such as time-of-day, day-of-week or season. We account for parameter uncertainties via an empirical Bayes approach. Additionally we scan for second order space-time clusters in the transition dynamics of agents moving within a city. The method simultaneously evaluates the existence and significance of primary source and secondary destination clusters. It enables the user to address the question: "given a location in time and space, which---if any---are the most likely circular/elliptical/rectangular regions that agents usually visit subsequently." We reduce computational complexity by applying a multi-scale grid to the geographic area scanned.
|
