Daniel B' Neill neillcs'cmu'edu

1
Daniel B. Neill (neill_at_cs.cmu.edu)
Assistant Professor of Information Systems
Teaching 95-796, Statistics for IT Managers
90-???, Artificial Intelligence Tools for Policy
Automatic detection of relevant patterns in
massive datasets
Developing new statistical and computational
methods for pattern detection.
Applying these methods to medicine, public
health, and homeland security.
Disease surveillance
Tumor detection
Network intrusion detection
Terrorist group detection
Water quality monitoring
Brain imaging
Game theory, evolution and learning
How do behaviors such as cooperation and
norm-following emerge in a population of
individuals (humans, companies, agents)?
2
Research Projects
Automatic detection of disease outbreaks using ED
visits and pharmacy drug sales
Detection of anomalous groups and patterns for
homeland security
Network intrusion detection
Terrorist group detection

Statistical methods for more accurate and useful
detection
More timely detection of emerging outbreaks, with
fewer false positives.
Better characterization of outbreaks (size,
location, severity, potential causes).
Computational methods for fast and scalable
detection in massive datasets
Fast spatial scan 100-1000x faster cluster
detection, no loss of accuracy.
3
Early detection of disease outbreaks can save
thousands of lives, and can significantly reduce
the human and economic impacts of an outbreak.
Our solution syndromic surveillance
Automatic monitoring of electronically available
public health data sources in near real-time.
(ED visits, OTC drug sales, 911 calls, etc.)
Automatic detection of anomalous patterns that
are indicative of an outbreak, using a variety of
new statistical methods.
Bioterrorist attacks, e.g. release of anthrax
spores
Automatic reporting of alerts to public health
officials via a web-based graphical interface.
We are currently performing surveillance on a
daily basis, monitoring data from over 20,000
stores and hospitals nationwide.
Avian influenza and other emerging infectious
diseases
4
A recent potential outbreak
Spike in sales of pediatric electrolytes near
Columbus, Ohio
GIS map view
Here is the potential outbreak.
Information about the region
Time series of potential outbreak region
5
A recent potential outbreak
Spike in sales of pediatric electrolytes near
Columbus, Ohio
Further investigation confirms GI emergency
department visits were also significantly
increased.
We believe that this increase in electrolyte
sales resulted from a small, localized GI
outbreak.
6
Interested? More details on my web page
http//www.cs.cmu.edu/neill Or e-mail me
at neill_at_cs.cmu.edu