HIERARCHICA

L CLUSTERING

Ronak Jangid – IU1841230019

T I T L E
Clustering is a technique that groups similar objects
such that the objects in the same group are more 2
similar to each other than the objects in the other
groups. The group of similar objects is called

M a y
a Cluster.

2 1 ,
2 0 X X
 P R E S E N TAT I O N
HIERARCHICAL CLUSTERING

• Hierarchical Clustering is an unsupervised

T I T L E
clustering algorithm which involves creating
clusters that have predominant ordering from
top to bottom.
• F o r e . g . : A l l fi l e s a n d f o l d e r s o n o u r h a r d d i s k
are organized in a hierarchy. 3

M AY
• This clustering technique is divided into two
types:

2 1 ,
• Agglomerative and Divisive

2 0 X X
 P R E S E N TAT I O N
A G G L O M E R AT I V E

T I T L E
It is the most common type of hierarchical clustering
which is used to group object based on clusters based 4
on their similarity.
It follows bottom up approach, i.e. each observation

M a y
starts in its own cluster, and pairs of cluster are

2 1 ,
merged as one moves up the hierarchy.

2 0 X X
 P R E S E N TAT I O N
HOW DOES IT WORK?

T I T L E
5

M a y
Step 1 Step 2 Step 3 Step 4

2 1 ,
Make each data point a single Take the two closest data points Take the two closest clusters and Repeat Step-3 until you are left

2 0 X X
point cluster – forms N cluster. and make them one cluster – make them one cluster - forms N- with only one cluster.
forms N-1 cluster. 2 cluster.
W H AT I S D E N D R O G R A M ?

P R E S E N TAT I O N
• A Dendrogram is a type of tree diagram
showing hierarchical relationships
b e t w e e n d i ff e r e n t s e t s o f d a t a .

T I T L E
6

M AY
2 1 ,
2 0 X X
 P R E S E N TAT I O N
PA R T S O F T H E D E N D R O G R A M

T I T L E
• The clades are the branch and are arranged to how similar (or
dissimilar) they are. The greater the difference in height, the
more dissimilarity.
• Each clade has one or more leaves. 7
• Leaves A, B, C are more similar to each other than leaves D, E, F.
• Leaves D, E are more similar to each other than A, B, C and F.

M a y
• And leaf F is substantially different from all the other leaves.

2 1 ,
2 0 X X
 P R E S E N TAT I O N
DIVISIVE CLUSTERING
• Divisive Hierarchical Clustering
starts with a single cluster
containing all the objects and then
split the cluster into two least

T I T L E
similar cluster based on their
characteristics.
• We p r o c e e d w i t h t h e s a m e p r o c e s s
until there is one cluster each 8
observation.
• It follows top-down approach.

M AY
2 1 ,
2 0 X X
DIFFERENT WAYS TO
MEASURE THE
DISTANC E B ETWEEN
TWO CLUSTERS
Types of linkage methods are:
1. Simple Linkage
2. Complete Linkage
3. Average Linkage
4. Centroid Linkage
 P R E S E N TAT I O N
SIMPLE LINKAGE

T I T L E
• Simple Linkage is also known as Minimum Linkage method.
• In this method, the distance of two clusters is defined as the
minimum distance between an object (point) in one cluster and
an object in another cluster. 10
• This method is also known as nearest neighbor method.

M a y
2 1 ,
2 0 X X
 P R E S E N TAT I O N
COMPLETE LINKAGE

T I T L E
• Complete Linkage is also known as Maximum Linkage method.
• In this method, the distance of two clusters is defined as the
maximum distance between an object (point) in one cluster and
an object in another cluster. 11
• This method is also known as furthest neighbor method.

M a y
2 1 ,
2 0 X X
 P R E S E N TAT I O N
AV E R A G E L I N K A G E

T I T L E
• In this method, the distance between the two clusters is the
average between each cluster’s point to every point in the other
cluster.
• This method is also known as unweighted pair group method 12
with arithmetic mean.

M a y
2 1 ,
2 0 X X
 P R E S E N TAT I O N
CENTROID LINKAGE

T I T L E
• In this method, the distance between the two sets or clusters is
the distance between two mean vectors of the sets (clusters).
• At each stage, we combine the two sets that have the smallest
centroid distance. In simple words, it is the distance between the 13
centroids of the two sets.

M a y
2 1 ,
2 0 X X
 A D VA N TA G E S

P R E S E N TAT I O N
• I t is e a s y t o im pl e m e n t.
• I t ca n p r o d u c e a n o r de r in g o f ob je c ts , w h ic h
m a y b e in f o r m at iv e f o r th e d i s p la y.

T I T L E
• Th e r e is n o n ee d t o p r e - s p e c if y t h e n u m be r o f
c lu s t er s .
• S m al le r c l u s te r s w ill b e c r e a te d , w h ic h m a y 14
d is c o v e r s im i la r i tie s in da ta .

M AY
2 1 ,
2 0 X X
 D I S A D VA N TA G E S

P R E S E N TAT I O N
• I t g iv e s b e s t r e s u lt s i n s o m e ca s es o n ly.
• Th e a lg o r ith m c a n n e v e r u n d o w h a t w a s d o n e
p r e v io u s ly, w h ic h m e an s if th e o b je c ts m a y

T I T L E
h a ve b e e n i nc o r r e c tly gr o u p e d a t an e a r lie r
s ta g e , a n d th e s a m e r e s u lt s h o ul d b e c lo s e to
e n s u r e it.
15
• S m al le r c l u s te r s w ill b e c r e a te d , w h ic h m a y
d is c o v e r s im i la r i tie s in da ta .

M AY
2 1 ,
2 0 X X
THANK YOU
16

IU1841230006 – Bharat Chauhan