Title: Modeling Count Data over Time Using Dynamic Bayesian Networks Jonathan Hutchins Advisors: Professor
1Modeling Count Data over Time Using Dynamic
Bayesian NetworksJonathan HutchinsAdvisors
Professor Ihler and Professor Smyth
2Sensor Measurements Reflect Dynamic Human Activity
Optical People Counter at a Building Entrance
Loop Sensors on Southern California Freeways
3Outline
- Introduction, problem description
- Probabilistic model
- Single sensor results
- Multiple sensor modeling
- Future Work
4Modeling Count Data
In a Poisson distribution mean variance ?
5Simulated Data
variance
mean people count
15 weeks, 336 time slots
6Building Data
variance
mean people count
7Freeway Data
variance
mean people count
8One Week of Freeway Observations
9(No Transcript)
10One Week of Freeway Data
11Detecting Unusual Events Baseline Method
Baseline model
car count
Unsupervised learning faces a chicken and egg
dilemma
12Persistent Events
Notion of Persistence missing from Baseline model
13Quantifying Event Popularity
Ideal model
Baseline model
14My contribution
- Adaptive event detection with time-varying
Poisson processes A. Ihler, J. Hutchins, and P.
Smyth Proceedings of the 12th ACM SIGKDD
Conference (KDD-06), August 2006. - Baseline method, Data sets, Ran experiments
- Validation
- Learning to detect events with Markov-modulated
Poisson processes A. Ihler, J. Hutchins, and P.
Smyth ACM Transactions on Knowledge Discovery
from Data, Dec 2007 - Extended the model to include a second event type
(low activity) - Poisson Assumption Testing
- Modeling Count Data From Multiple Sensors A
Building Occupancy Model J. Hutchins, A. Ihler,
and P. Smyth - IEEE CAMSAP 2007,Computational Advances in
Multi-Sensor Adaptive Processing, December 2007.
15Graphical Models
"Graphical models are a marriage between
probability theory and graph theory. They provide
a natural tool for dealing with two problems that
occur throughout applied mathematics and
engineering -- uncertainty and complexity
Michael Jordan 1998
16Directed Graphical Models
hidden
Observed Count
observed
Event
Rate Parameter
17Directed Graphical Models
- Nodes ? variables
- Edges ? direct dependencies
18Graphical Models Modularity
Observed Countt
Observed Countt-2
Observed Countt-1
Observed Countt2
Observed Countt1
19Graphical Models Modularity
Poisson Rate ?(t)
Day, Timet-1
Day, Timet
Day, Timet1
Normal Countt-1
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
Observed Countt1
20Graphical Models Modularity
21Graphical Models Modularity
Poisson Rate ?(t)
Day, Timet-1
Day, Timet
Day, Timet1
Normal Countt-1
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
Observed Countt1
Eventt
Eventt-1
Eventt1
22Graphical Models Modularity
Eventt
Eventt-1
Eventt1
Event State Transition Matrix
23Poisson Rate ?(t)
Day, Timet-1
Day, Timet
Day, Timet1
Normal Countt-1
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
Observed Countt1
Event Countt
Event Countt-1
Event Countt1
Eventt
Eventt-1
Eventt1
Event State Transition Matrix
24a
Poisson Rate ?(t)
Day, Timet-1
Day, Timet
Day, Timet1
Normal Countt-1
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
Observed Countt1
Event Countt
Event Countt-1
Event Countt1
Eventt
Eventt-1
Eventt1
?
?
?
Event State Transition Matrix
ß
25Poisson Rate ?(t)
Day, Timet-1
Day, Timet
Day, Timet1
Normal Countt-1
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
Observed Countt1
Event Countt
Event Countt-1
Event Countt1
Eventt
Eventt-1
Eventt1
Event State Transition Matrix
Markov Modulated Poisson Process (MMPP) model
e.g., see Heffes and Lucantoni (1994), Scott
(1998)
26Approximate Inference
27Gibbs Sampling
28Gibbs Sampling
y
x
29Block Sampling
30Gibbs Sampling
Poisson Rate ?(t)
Day, Timet-1
Day, Timet
Day, Timet1
Normal Countt-1
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
Observed Countt1
Event Countt
Event Countt-1
Event Countt1
Eventt
Eventt-1
Eventt1
Event State Transition Matrix
31Gibbs Sampling
Poisson Rate ?(t)
Poisson Rate ?(t)
Poisson Rate ?(t)
For the ternary valued event variable with
chain length of 64,000 Brute
force complexity
Day, Timet-1
Day, Timet
Day, Timet1
Normal Countt-1
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
Observed Countt1
Event Countt
Event Countt-1
Event Countt1
Eventt
Eventt-1
Eventt1
Event State Transition Matrix
Event State Transition Matrix
Event State Transition Matrix
32Gibbs Sampling
Poisson Rate ?(t)
Poisson Rate ?(t)
Day, Timet-1
Day, Timet-1
Observed Countt-1
Observed Countt-1
Event Countt-1
Event Countt-1
Normal Countt-1
Normal Countt-1
Eventt
Eventt-1
Eventt1
A
A
A
33(No Transcript)
34Chicken/Egg Delima
car count
35Event Popularity
car count
car count
36Persistent Event
Notion of Persistence missing from Baseline model
37Persistent Event
38Detecting Real Events Baseball Games
Remember the model training is completely
unsupervised, no ground truth is given to the
model
39Multi-sensor Occupancy Model
Modeling Count Data From Multiple Sensors A
Building Occupancy Model J. Hutchins, A. Ihler,
and P. Smyth IEEE CAMSAP 2007,Computational
Advances in Multi-Sensor Adaptive Processing,
December 2007
40Where are the People?
Building Level
City Level
41Sensor Measurements Reflect Dynamic Human Activity
Optical People Counter at a Building Entrance
Loop Sensors on Southern California Freeways
42Application Multi-sensor Occupancy Model
CalIt2 Building, UC Irvine campus
43Building Occupancy, Raw Measurements
Occt Occt-1 inCountst-1,t outCountst-1,t
44Building Occupancy Raw Measurements
Under-counting
Noisy sensors make raw measurements of little
value
45Adding Noise Model
Poisson Rate ?(t)
Day, Timet-1
Day, Timet
Normal Countt-1
Normal Countt-1
Observed Countt
Observed Countt-1
True Countt-1
True Countt
Event Countt
Event Countt-1
Eventt
Eventt-1
Event State Transition Matrix
46Probabilistic Occupancy Model
Time t
Time t1
Constraint
Occupancy
Occupancy
Out(Exit) Sensors
Out(Exit) Sensors
In(Entrance) Sensors
In(Entrance) Sensors
4724 hour constraint
Geometric Distribution, p0.5
Constraint
? ? ?
Occupancy
? ? ?
? ? ?
Building Occupancy
47
48Learning and Inference
- Gibbs Sampling Forward-Backward Complexity
Occupancy
Occupancy
Out(Exit) Sensors
Out(Exit) Sensors
In(Entrance) Sensors
In(Entrance) Sensors
49Typical Days
Building Occupancy
Thursday Friday
Saturday
50Missing Data
Building Occupancy
time
51Corrupted Data
Building Occupancy
Thursday
Friday
52Future Work
- Freeway Traffic
- On and Off ramps
- 2300 sensors
- 6 months of
- measurements
53Sensor Failure Extension
54Spatial Correlation
55Four Off-Ramps
56Publications
- Modeling Count Data From Multiple Sensors A
Building Occupancy Model J. Hutchins, A. Ihler,
and P. Smyth - IEEE CAMSAP 2007,Computational Advances in
Multi-Sensor Adaptive Processing, December 2007. - Learning to detect events with Markov-modulated
Poisson processes A. Ihler, J. Hutchins, and P.
Smyth ACM Transactions on Knowledge Discovery
from Data, Dec 2007 - Adaptive event detection with time-varying
Poisson processes A. Ihler, J. Hutchins, and P.
Smyth Proceedings of the 12th ACM SIGKDD
Conference (KDD-06), August 2006. - Prediction and ranking algorithms for event-based
network data - J. O Madadhain, J. Hutchins, P. Smyth
- ACM SIGKDD Explorations Special Issue on Link
Mining, 7(2), 23-30, December 2005