Automatic Inference of Anomalous Events from California Traffic Patterns

1
Automatic Inference of Anomalous Events from
(California) Traffic Patterns

• Sean Li, Electrical Engineering. Professor Dr.
  Padhraic Smyth

Previous Work
Surf-IT Project
This Surf-IT project built on work presented in
Adaptive event detection with time-varying
Poisson processes A. Ihler, J. Hutchins, and P.
Smyth, Proceedings of the 12th ACM SIGKDD
Conference (KDD-06), to appear, 2006
Real Time Traffic Event Detection In this
project, we developed a web-based system that
automatically identifies anomalous events on a
freeway by analyzing traffic pattern data from
sensors. By implementing the time varying Poisson
model, this system is capable of detecting any
unexpected events in any given location, day,
time etc. This system can display both the
real-time traffic data and "toggle" to a display
what the model considers to be unusual.
Time Series Count Data
Website shows map and displays predictions for
the last 2 hours
4
2
Markov Modulated Poisson Process
Baseline Model has Limitations
1
The Chicken and Egg problem
Graphical model for event process (z(t)) and
observed counts (N(t))
Illustration Of The Real Time Traffic Event
Detection System (1)Traffic data stored in the
PEMS (Freeway Performance Measurement System) FTP
server. (2) Perl script receives
/extracts/parts/stores data into Mysql data
base. (3) C inference code implements the
time-varying Poisson model and returns the
calculated event probability in real time. (4)
Ruby on Rails and JAVA software is used to create
and update the web-based system to reflect the
current freeway condition.
Graphical model for Normal Counts (No(t)) and
the Poisson rate parameter
False Positives, Persistence and Duration
Baseline model-lower threshold
Baseline model
3
ACKNOWLEDGEMENTS This project was conducted under
the guidance of Dr. Padhraic Smyth and Jon
Hutchins. PEMS Data collecting is done by EECS
department at UC Berkeley.
S ummer U ndergraduate 2 R esearch
0 F ellowship in 0 I
nformation 6 T echnology
shuangl_at_uci.edu www.research.calit2.net/studen
ts/surf-it2006 www.calit2.net
2
(No Transcript)