|
Activity Number:
|
98
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statisticians in Defense and National Security
|
| Abstract - #307593 |
|
Title:
|
Geospatial Modeling in an Information Theoretic Framework as Applied to Forecasting of Insurgent Activity
|
|
Author(s):
|
Jason Dalton*+
|
|
Companies:
|
Spatial Data Analytics Corporation
|
|
Address:
|
1950 Old Gallows Road, Vienna, VA, 22182,
|
|
Keywords:
|
spatial ; forecast ; mutual information ; area reduction ; resource allocation
|
|
Abstract:
|
Geospatial modeling for defense and intelligence amounts to providing an ordered set of location elements where the order conveys a difference in likelihood for the occurrence of a particular event of interest. This presentation will describe one approach to geospatial modeling that accounts for redundant and irrelevant variables through a mutual information measure applied to pair-wise comparisons of features. By using distribution divergence measures, derived spatial features can be removed when they do not contribute positively to the resulting forecast. A chief innovation of this work is its use of a credit/blame assignment to variables to adjust weights over subsequent assessments. This work builds on prior work by developing a utility model for target decision tradeoffs. Applications in IED emplacement, high-value individual tracking, and buried targets will be discussed.
|
