UAegean, FME: Special Lecture

nervousripSecurity

Nov 5, 2013 (4 years and 8 days ago)

99 views

Stelios Lelis

UAegean, FME: Special Lecture


Social Media & Social Networks
(SM&SN)


http://www.youtube.com/watch?v=6a_KF7TYKVc


social media (recap)


Offer means for people to communicate that
complements

face
-
to
-
face meeting.


Offer ways for people to
broadcast

information to larger groups


Offer ways for people to make social information
persistent


Offer ways for people to make
ego
-
alter

communication and
to see alter
-
alter communication



Offer ways to form
community around interests
, e.g. music,
and also ‘long
-
tail’ interests, e.g. train spotting


Offer ways to bridge constraints of
time and place


May also
enrich same
-
time
,
same place social interaction



Social Network


A
social network

is a set of people or groups of people with
some pattern of contacts or interactions between them


Social network analysis focuses on the
relations among
people
, and not individual people and their attributes


The social network is a group of people which we call
nodes,
and connections between them called

edges

(or
ties
)

Node

Tie

Ties


Different types of ties: family, friend, personal / professional, ego
-
perceived / alter
-
perceived / mutual


Directed (Flickr) / Undirected (Facebook)


Strong & Weak ties


Amount of time, emotional intensity, intimacy and reciprocal services

Strong
Tie

Weak
Tie

Path length & Neighbourhoods


Path length
:

number of edges in the shortest path between two nodes


k
-
hop neighbourhood

of a node: the set of nodes that can be reached
through paths of length
k
(friends… and friends of friends… etc.)

1
-
hop

2
-
hop

3
-
hop

4
-
hop

Small
-
world


Most pairs of nodes seem to be connected by a short path through
the network (
Six degrees

of separation)


Average

path length

(L):
Mean path length between nodes in the
network


Diameter
(
D
): Maximum path length between nodes in the network


Small
-
world implies that spread of information will be fast

L

D

Flickr

5.67

27

Live Journal

5.88

20

Orkut

4.25

9

YouTube

5.10

21

It’s a small
world after all

Clustering


Friends of friends are likely to be friends


Clustering coefficient, C

(0


C



1)


Density of triangles in the network


Density of links that exist between one’s friends

C

Flickr

0.313

Live Journal

0.330

Orkut

0.171

YouTube

0.136

Degree distribution


Degree of a node: the number of edges connected to a node


Degree, out
-
degree & in
-
degree


Most nodes have few edges while few nodes have many edges
(Scale
-
free, power law degree distribution)

Flickr

Node
degree = 4


Mixing patterns


Homophily & Assortativity


Homophily

(or
assortative mixing
): The tendency of people to associate
and connect with similar others


Mixing by lines of interest, occupation, age, race, etc.


Assortativity: The likelihood of nodes to connect to other nodes with
similar degrees
(high degree to high degree, forming a core)


Social networks are assortative


Important for the flow of information

s

r

Flickr

0.49

0.20

Live Journal

0.34

0.18

Orkut

0.36

0.07

YouTube

0.19

-
0.03

Community structure


Groups of people in the network that have a high density of connections
within them and a lower density of connections between them

Friendship
network of
children in a
US school

Structural holes


The weaker connections between groups


A structural hole between two groups does not mean that people in
the groups are unaware of one another.


It only means that people are focused on their own activities such
that they do not attend to the activities of people in the other group.

Information Propagation in the
Flickr Social Network

1.
Comments

2.
Notes

3.
Favourites

4.
People

5.
Tags



Connect to friends


Join groups

Information propagation & Data Collection


Information propagation:
photos’ favourite
-
marking


Friends
: users in the contact list of a user


Fans
: users who include a photo in their favourite photos


Data Collection


Crawl of the social network graph once per day for 104 days


Each user’s favorite photos


Timestamp when each photo was favourite
-
marked

Local vs. Global Picture Popularity


Different pictures are popular among the different social network
regions;


compare global and local
hotlists (top 100 pictures)


no overlap between
1
-
hop and global


overlap increases as
neig/hoods get wider


4
-
hop neig/hood covers
36% of entire graph
(small
-
world)

Distance from fans to uploaders


Strong locality across all popularity levels


Propagation is limited and photos rarely spread beyond the
immediate vicinity of their uploaders

Patterns of popularity growth

Active growth

Surge
-
increase

Sluggish

Growth evolves differently but shares common patterns

Influenced
by node
in
-
degree

External event /
High reproduction rate

Long
-
term trends in popularity growth


Flickr users take a long period of
time to learn about interesting
pictures


Popular photos show an active
rise in popularity during the first
few days, and then enter a period
of steady linear growth


Less popular photos attract their
limited fan population early on
during their lifetime and then they
become dormant


Information propagation via social links


Social cascade: Information (or decisions/habits) spreading
through a social network one
-
hop at a time


Social cascade plays a significant role in propagating
information

… for both popular and unpopular pictures

Social cascades and popularity


Social cascades play an important role in picture popularity


Uploaders play a crucial role in the social cascades of less popular
pictures


Social cascades of popular pictures spread information beyond the
immediate vicinity of the uploader

Peer pressure


The probability of a user becoming a fan of a photo increases
with the number of friends who are already fans of the photo

The power of social networks


Information does spread through social connections


Behaviours, habits, traits & biological indices spread alike?


The case of obesity…

http://www.nejm.org/doi/full/10.1056/NEJMsa066082


Summary


Measures of network structure: path length, diameter, clustering,
node degree


Properties of social networks: homophily, small
-
world, local
clustering, assortativity, community structure


Information Propagation Examples


Information propagates through social connections


More important for popular pictures


Most information does not spread out through the entire Flickr
network…


…but traits do in other networks (e.g. Heart Study Obesity Network)