Extract Agent-based Model from Communication Network
description
Transcript of Extract Agent-based Model from Communication Network
![Page 1: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/1.jpg)
Extract Agent-based Model from
Communication Network
Hung-Ching (Justin) ChenMatthew Francisco
Malik Magdon-IsmailMark Goldberg
William WallanceRPI
![Page 2: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/2.jpg)
Goal
n Deduce something about “nature” of the society:n e.g., Do actors generally have a propensity to
join small groups or large groups?n Predict the society’s future:
n e.g., How many social groups are there after 3 months?
n e.g., What is the distribution of group size?
Given a society’s communication history,can we:
![Page 3: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/3.jpg)
General Approach
Society’s History
Society’s Future
“Predict”(Simulate)
“Learn” IndividualBehavior
(Micro-Laws)
![Page 4: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/4.jpg)
General Approach
Society’s History
Society’s Future
“Predict”(Simulate)
“Learn” IndividualBehavior
(Micro-Laws)
![Page 5: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/5.jpg)
Social Networks• Individuals (Actors)
• Groups
12
3
![Page 6: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/6.jpg)
Social Networks• Individuals (Actors)
• Groups
1 2
3
- Join - Leave
![Page 7: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/7.jpg)
4
Social Networks• Individuals (Actors)
• Groups
1
3
- Join - Leave
- Disappear - Appear
2
- Re-appear
![Page 8: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/8.jpg)
Society’s History
![Page 9: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/9.jpg)
General Approach
Society’s History
Society’s Future
“Predict”(Simulate)
“Learn” IndividualBehavior
(Micro-Laws)
![Page 10: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/10.jpg)
Modeling of Dynamics
Micro-Law# 1
Micro-Law# 2
Micro-Law# N
…
Parameters HistoryGroups & Individuals
Actions Join / Leave / Do Nothing
![Page 11: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/11.jpg)
Example of Micro-Law
Actor X likes to join groups.
Parameter
SMALLLARGE
![Page 12: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/12.jpg)
ViSAGEVirtual Simulation and Analysis of Group
Evolution
Real Action
ActorChoice
State: Properties of Actors and Groups
Decide Actors’ Action
Process Actors’ Action
Feedbackto Actors
State
StateState update
NormativeAction
State
![Page 13: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/13.jpg)
General Approach
Society’s History
Society’s Future
“Predict”(Simulate)
“Learn” IndividualBehavior
(Micro-Laws)
![Page 14: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/14.jpg)
Learning
Learn
Parameters #1in
Micro-Laws?
?
Communications
Parameters #2in
Micro-Laws
![Page 15: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/15.jpg)
Groups & Group Evolution
Communications
Groups: Overlappingclustering
GroupsEvolution
Groupevolution: Matching
![Page 16: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/16.jpg)
Actor’s Typesn Leader: prefer small group size and is most
ambitiousn Socialite: prefer medium group size and is
medium ambitiousn Follower: prefer large group size and is
least ambitious
![Page 17: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/17.jpg)
Learning Actors’ Type
n Maximum log-likelihood learning algorithmn Cluster algorithmn EM algorithm
![Page 18: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/18.jpg)
Testing Simulation Data
![Page 19: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/19.jpg)
Testing Real DataCluster
AlgorithmLearned Actors’ Types
Leader Socialite FollowerNumber of Actor 822 550 156
Percentage 53.8% 36.0% 10.2%
EMAlgorithm
Learned Actors’ TypesLeader Socialite Follower
Number of Actor 532 368 628Percentage 34.8% 24.1% 41.1%
![Page 20: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/20.jpg)
General Approach
Society’s History
Society’s Future
“Predict”(Simulate)
“Learn” IndividualBehavior
(Micro-Laws)
![Page 21: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/21.jpg)
Testing & Simulations
Micro-Laws&
Parameters# 1
Simulate
Micro-Laws&
Parameters# 2
Simulate
![Page 22: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/22.jpg)
Prediction
![Page 23: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/23.jpg)
Prediction
![Page 24: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/24.jpg)
Future Work
n Test Other Predictionsn e.g., membership in a particular group
n Learn from Other Real Datan e.g., emails and blogs
![Page 25: Extract Agent-based Model from Communication Network](https://reader035.fdocuments.in/reader035/viewer/2022070500/5681684c550346895dde4915/html5/thumbnails/25.jpg)
Questions?