Home // GEOProcessing 2013, The Fifth International Conference on Advanced Geographic Information Systems, Applications, and Services // View article

Simple Methods for Reasoning about Behavior Patterns on Graphs Given Extremely Sparse Observations

Authors:
R. Paul Wiegand
Steven Prager

Keywords: parse observations; fair paths; path analysis

Abstract:
We consider the situation where fixed observations of moving entities are sparse and the goal is to learn as much as possible about their patterns of activity, before and after such observations (e.g, cameras at a few intersections in a city). Here we present a method for estimating probable paths within a network given a limited set of vertex observations and limited a priori assumptions about individual entity behavior. We divide the process of analysis into two phases: a learning phase in which aggregate information about many entities is obtained and used to construct simple models given potential observations, and a reasoning phase in which resampling methods produce probable paths a specific entity may have taken. To accomplish this, we extend a fair and efficient method for randomly selecting unconditioned paths within a network in order to draw paths conditioned on limited, partial observations. The methods are validated by analyzing hypothetical observations of entities moving on an existing city street network. Our results show the scaling properties of this approach by optimizing the locations of different numbers of fixed potential observation points to obtain as maximal coverage of the area as possible. We then construct a variety of models based on an extremely sparse observational scenario and demonstrate quantitatively and visually that these simple methods, combined with structural information inherent in the graph itself, can provide a great deal of context information about an individual entity’s possible movement patterns.

Pages: 73 to 80

Copyright: Copyright (c) IARIA, 2013

Publication date: February 24, 2013

Published in: conference

ISSN: 2308-393X

ISBN: 978-1-61208-251-6

Location: Nice, France

Dates: from February 24, 2013 to March 1, 2013