Spread of Information in a Social Network Using Influential Nodes

24
Spread of Information in a Social Network Using Influential Nodes Arpan Chaudhury Partha Basuchowdhuri Subhashis Majumder (Heritage Institute of Technology, Kolkata)

Transcript of Spread of Information in a Social Network Using Influential Nodes

Page 1: Spread of Information in a Social Network Using Influential Nodes

Spread of Information in a

Social Network Using

Influential Nodes

Arpan Chaudhury

Partha Basuchowdhuri

Subhashis Majumder

(Heritage Institute of

Technology, Kolkata)

Page 2: Spread of Information in a Social Network Using Influential Nodes

Introduction A social network is a set of actors

that may have relationships with one another.

It can be further described as the graph of relationships and interactions between a set of individuals.

Nodes are the individuals (visitors) within the networks, and ties are the different types of relationships between the individuals.

Page 3: Spread of Information in a Social Network Using Influential Nodes

Social Networks

In a social network

represents a node

or a person

represents an edge

or a relationship

Page 4: Spread of Information in a Social Network Using Influential Nodes

Our Aim

To find the set

of the most

“influential” such

that we can

maximize the

spread of

information

through those

“influentials”

Page 5: Spread of Information in a Social Network Using Influential Nodes

Where to use….

-Viral marketing

-Word-of-mouth

marketing

strategies

Strategies to stop-

-Telecom churn

-Disease spread

Page 6: Spread of Information in a Social Network Using Influential Nodes

Motivation behind our approach

- lesser the degree

of a node, the

correlation plot

shows that if we

choose that node as

that node as the

only seed to start

the spread of

information in the

network, more the

# of hops it takes

Page 7: Spread of Information in a Social Network Using Influential Nodes

Why MST (Max. Spanning Tree) ?

Path through which information is more likely to flow

Identify nodes with high spread potential

Identify bridges passing information from one group to another

Remove insignificant edges retaining all nodes

Page 8: Spread of Information in a Social Network Using Influential Nodes

0

4

2

1

5

6

7

3

8

3.5

2.5

4.5

3

44

4.5

3.54

2.5

2.53

3.5

D(0)=4D(1)=1

D(2)=2

D(3)=5D(4)=3

D(7)=3

D(5)=3

D(6)=4

D(8)=1

w(x) denotes weight of an edge x.

c(x) denotes cost of an edge x.

0.4

0.33

0.250.22

Max. Spanning Tree of the Network

Weight of the EdgesDegree of the NodesA Network

Edges in maximum spanning tree of the network represent the most

probable path of information flow

0.25

0.25

0.4

0.22

Modifying the graph to build MST

)(

1)(

2

)()()(

ij

ij

ij

eweC

jDiDew

Page 9: Spread of Information in a Social Network Using Influential Nodes

Information Flow Path

Influential

Node

Information

Flow Path

Page 10: Spread of Information in a Social Network Using Influential Nodes

Algorithm – Part I

Page 11: Spread of Information in a Social Network Using Influential Nodes

Algorithm – Part II

Page 12: Spread of Information in a Social Network Using Influential Nodes

Experimental Results

Page 13: Spread of Information in a Social Network Using Influential Nodes

Dolphin Network

Dolphin network

with 62 vertices and

159 edges.

Popularly used as

benchmark data for

community

detection algorithms

in SNA.

Page 14: Spread of Information in a Social Network Using Influential Nodes

Dolphin Network

Maximum spanning

tree of dolphin

network.

Page 15: Spread of Information in a Social Network Using Influential Nodes

Dolphin Network

Core of dolphin

network, with k=7

Page 16: Spread of Information in a Social Network Using Influential Nodes

AS Relationship Network

AS relation network

with 6474 vertices

and 13895 edges

from CAIDA.

Used here to test

our algorithm for a

large dataset

Page 17: Spread of Information in a Social Network Using Influential Nodes

Maximum spanning

tree (MST) of

AS relationship

network.

AS Relationship Network

Page 18: Spread of Information in a Social Network Using Influential Nodes

Core of AS

relationship network,

with k=18 (dth=50)

AS Relationship Network

Page 19: Spread of Information in a Social Network Using Influential Nodes

Greedy k-center Vs. core-finding

Green/ cyan – first hop, Yellow – second hop,

Blue – third hop, Pink – fourth hop, White – fifth hop

Page 20: Spread of Information in a Social Network Using Influential Nodes

Comparative study

Page 21: Spread of Information in a Social Network Using Influential Nodes

Hop by hop spread comparison

Page 22: Spread of Information in a Social Network Using Influential Nodes

Hop by hop spread comparison

Page 23: Spread of Information in a Social Network Using Influential Nodes

Conclusion

Efficient and accurate compared to k-center.

Spread is simplistic and not community based,

hence takes very less time.

Work has been updated based on degree

discount and the model has been generalized

according to independent cascade model (all

edges won’t lead to spread).

Page 24: Spread of Information in a Social Network Using Influential Nodes

Thank You