Transcript PPT

Mining and Visualizing the
Evolution of Subgroups in
Social Networks
Falkowsky, T., Bartelheimer, J. & Spiliopoulou,
M. (2006) IEEE/WIC/ACM International
Conference on Web Intelligence, pp. 52-58
Presented by Danielle Lee
Outline





Problem
Research Purpose
Data Set
First Approach : Statistical Analyses and
Visualization for relatively stable
communities
Second Approach : Detection of the
subgroup evolution in high fluctuating
communities
Problem



A community has rather stable structure
with a small amount of fluctuating
members and they participate in over a
long time.
Another community has high dynamic
structure whose members and their
networks keep changing over time.
Different community detection and
visualization methods are needed.
Research Purpose

To propose statistical method and
visualization to analyze the
formation of subgroups and the
timely change of online communities
on the level of sub-groups
Data Set



Taken from an online international
student community in the University
of Magdeburg.
About 1000 members from more
than 50 countries
250,000 guestbook entries over a
period of 18 months
Evolution of Subgroups in
Static Structure (Contd.)

Mining for subgroups in Social
Networks
1)
2)
3)
Partitioning data by time axis
Weight graph Gt of interactions
between individuals for each time
windows is built.
Hierarchical edge betweenness
clustering of the graph is applied in
each time window
Evolution of Subgroups in
Static Structure (Contd.)
Subgroups
Communication
within one
community
Detailed
information
at a certain
time point
time
Evolution of Subgroups in
Static Structure (Contd.)

Analyzing Subgroup Dynamics

Track a detected subgroup over time by measuring
the structural equivalence






Stability
Density and cohesion
Euclidean distance
Correlation coefficient
Group activity
The measures are computed for each time window


Fixed : A chosen time window is compared with all other
windows
Periodical : Each time window is compared to the
previous time window
Evolution of Subgroups in
Static Structure
Kinds of
Measurement
Each
Subgroup
Dynamics of Communities with
Fluctuating Members (contd.)

Clustering subgroups as a community



Establish a graph of subgroups to denote similarity about them
Similarity have been discovered as the overlap of members
between two subgroups
Two subgroups are similar if their overlap exceeds a given
threshold.
Dynamics of Communities with
Fluctuating Members (contd.)

Visualizing the Evolution of Subgroups
Community
Clustering
Control
Panel
Dynamics of Communities with
Fluctuating Members (contd.)

Community History View
Dynamics of Communities
with Fluctuating Members
Thank you