Scribd
Upload
21 views21 pages

2 Lecture 6

This document discusses triadic closure and methods for measuring clustering and closure in networks. It introduces two approaches for calculating global clustering coefficients: (1) calculating the percentage of triangles in the network and (2) calculating transitivity, which is the percentage of open triads that are closed triangles. These two approaches can yield different results as transitivity weights high degree nodes more heavily and considers network-wide closure, while local clustering coefficients consider closure from each node's perspective.

Uploaded by

PC Bhaskar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
21 views21 pages

2 Lecture 6

This document discusses triadic closure and methods for measuring clustering and closure in networks. It introduces two approaches for calculating global clustering coefficients: (1) calculating the percentage of triangles in the network and (2) calculating transitivity, which is the percentage of open triads that are closed triangles. These two approaches can yield different results as transitivity weights high degree nodes more heavily and considers network-wide closure, while local clustering coefficients consider closure from each node's perspective.

Uploaded by

PC Bhaskar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 21

Closure(Triadic Closure)

Why?
• Some times we need to know that what relations are more likely to
form in future.
• So we need to find some connected groups.
• Measurement of their connected ness will be studied in closure.
So triadic closure says those edges make the
triangles in network are more likely to be connected
So CD, HF, HD, and FB are edges that forming triangles hence termed
as triadic.

So Next question is can we measure the prevalence of triadic closure.


Can we levitate this concept from static to dynamic graph.
Clustering is the measurement of triadic
closure
• Lets measure the clustering from the point of view of a single node.
• It is known as local clustering coefficient.
Lets say you wanna to compute clustering
coefficient of Node C
Let’s do another example
Another example:
So when definition does not work then we take above assumption
How to do it by networkX
Global clustering coefficient
• Previous way figure out the local clustering coefficient.
• It make triadic closure locally in graph.
• Lets take care of global clustering coefficient for triadic closure.
• So we will see global clustering coefficient for entire network than
local coefficient for one local cluster.
• For this we follow two different approaches.
Approach-1
Approach-2
• In second approach we try to find the % of open triads which are
triangles

Triangles are three nodes connected with three edges

Triads are three nodes connected with two edges

Each triangle can have three different triads


Transitivity:
• So when you see one triangle you should see three triads also .
• With this we introduce second approach to find the global clustering
coefficient .i.e. transitivity.
i.e,
Triangles

This formula came b’cos one triangle = three triads


Transitivity is percentage of open triads that
are triangles or closed triads
• You can use networkX to get transitivity by a function of same name
as:
Bit discussion: (Comparisons)
• We have two different ways of calculating clustering coefficient.
• Are these methods are same, or different.
• If different then what is the diffrence?

-In this ring, Most nodes have


high LCC.
- Node at perimeter has LCC =1
since all neighbours are
connected
- Central nodes has less LCC
since many of it’s connections
are not connected
- So average LCC is very high i.e.
= 0.93
- Transitivity is 0.23 which is
Cont.
• Transitivity weights high degree nodes higher and penalizes more that
resulted in low transitivity.
• Now take a network where each node has very low LCC.
-In this ring, Outer most
nodes(15) have LCC = 0
- Nodes in perimeter(5) has
high LLC
- So average LCC is very Low i.e.
= 0.23
- Transitivity is high 0.86 which
is very high
Conclusion:

• If high degree nodes are more ten LCC may high and transitivity low.
• If Low degree nodes are more ten LCC may low and transitivity high.

You might also like