Clustering

Lecture 1: Basics

Jing Gao
SUNY Buffalo

1
 Class Structure

• Topics
– Clustering, Classification
– Network mining
– Anomaly detection
• Expectation
– Sign-in
– Take quiz in class
– Two more projects on clustering and classification
– One more homework on network mining or anomaly detection
• Website
– http://www.cse.buffalo.edu/~jing/cse601/fa12/

2
 Outline
• Basics
– Motivation, definition, evaluation
• Methods
– Partitional,
– Hierarchical
– Density-based
– Mixture model
– Spectral methods
• Advanced topics
– Clustering ensemble
– Clustering in MapReduce
– Semi-supervised clustering, subspace clustering, co-clustering,
etc.

3
 Readings

• Tan, Steinbach, Kumar, Chapters 8 and 9.

• Han, Kamber, Pei. Data Mining: Concepts and Techniques.
Chapters 10 and 11.
• Additional readings posted on website

4
 Clustering Basics

• Definition and Motivation

• Data Preprocessing and Similarity Computation
• Objective of Clustering
• Clustering Evaluation

5
 Clustering
• Finding groups of objects such that the objects in a group will
be similar (or related) to one another and different from (or
unrelated to) the objects in other groups

Inter-cluster
Intra-cluster distances are
distances are maximized
minimized

6
 Application Examples

• A stand-alone tool: explore data distribution

• A preprocessing step for other algorithms
• Pattern recognition, spatial data analysis, image processing,
market research, WWW, …
– Cluster documents
– Cluster web log data to discover groups of similar access
patterns

7
 Clustering Co-expressed Genes
Gene Expression Data Matrix Gene Expression Patterns

Co-expressed Genes

Why looking for co-expressed genes?

 Co-expression indicates co-function;
 Co-expression also indicates co-regulation.

8
 Gene-based Clustering
1.5

0.5

Expression Value
0

-0.5

-1

-1.5
Time Point

1.5

0.5

Expression Level
0

-0.5

-1

-1.5
Time Points

1.5

0.5

Expression Value
0

-0.5

-1

-1.5
Time Points

Examples of co-expressed genes and coherent

patterns in gene expression data
Iyer’s data [2]

 [2] Iyer, V.R. et al. The transcriptional program in the response of human fibroblasts to serum. Science,
283:83–87, 1999.
9
 Other Applications

• Land use: Identification of areas of similar land use in an earth

observation database
• Marketing: Help marketers discover distinct groups in their
customer bases, and then use this knowledge to develop
targeted marketing programs
• City-planning: Identifying groups of houses according to their
house type, value, and geographical location
• Climate: understanding earth climate, find patterns of
atmospheric and ocean

10
 Two Important Aspects

• Properties of input data

– Define the similarity or dissimilarity between points
• Requirement of clustering
– Define the objective and methodology

11
 Clustering Basics

• Definition and Motivation

• Data Preprocessing and Distance computation
• Objective of Clustering
• Clustering Evaluation

12
 Data Representation

• Data: Collection of data objects Attributes

and their attributes

• An attribute is a property or Tid Refund Marital Taxable

Status Income Cheat
characteristic of an object
1 Yes Single 125K No
– Examples: eye color of a person,
temperature, etc. 2 No Married 100K No
3 No Single 70K
– Attribute is also known as No

dimension, variable, field, 4 Yes Married 120K No

characteristic, or feature 5 No Divorced 95K Yes
Objects
6 No Married 60K No

• A collection of attributes describe 7 Yes Divorced 220K No

an object 8 No Single 85K Yes

9 No Married 75K No
– Object is also known as record,
point, case, sample, entity, or 10
10 No Single 90K Yes
instance

13
 Data Matrix

• Represents n objects with p variables

– An n by p matrix
The value of the i-th
object on the f-th
Attributes attribute

 x11  x1 f  x 
1p 

Objects
      
x  x  x 
 i1 if ip 
      
 
 xn1  xnf  x 
np 

14
 Gene Expression Data

sample 1 sample 2 sample 3 sample 4 sample …

• Clustering genes
gene 1 0.13 0.72 0.1 0.57 •Genes are objects
gene 2 0.34 1.58 1.05 1.15 •Experiment conditions are
gene 3 0.43 1.1 0.97 1 attributes
gene 4 1.22 0.97 1 0.85
• Find genes with similar
gene 5 -0.89 1.21 1.29 1.08 function
gene 6 1.1 1.45 1.44 1.12
gene 7 0.83 1.15 1.1 1
gene 8 0.87 1.32 1.35 1.13
gene 9 -0.33 1.01 1.38 1.21
gene 10 0.10 0.85 1.03 1
gene
…

15
 Similarity and Dissimilarity

• Similarity
– Numerical measure of how alike two data objects are
– Is higher when objects are more alike
– Often falls in the range [0,1]
• Dissimilarity
– Numerical measure of how different are two data
objects
– Lower when objects are more alike
– Minimum dissimilarity is often 0
– Upper limit varies

16
 Distance Matrix

• Represents pairwise distance in n objects

– An n by n matrix
– d(i,j): distance or dissimilarity between objects i and j
– Nonnegative
– Close to 0: similar

 0 
d (2,1) 0 
 
d (3,1) d (3,2) 0 
 
    
d (n,1) d (n,2)   0
17
 Data Matrix -> Distance Matrix

s1 s2 s3 s4 …
g1 0.13 0.72 0.1 0.57
g1 g2 g3 g4 …
g2 0.34 1.58 1.05 1.15
g3 0.43 1.1 0.97 1 g1 0 d(1,2) d(1,3) d(1,4)
g4 1.22 0.97 1 0.85 g2 0 d(2,3) d(2,4)
g5 -0.89 1.21 1.29 1.08
g6 1.1 1.45 1.44 1.12 g3 0 d(3,4)
g7 0.83 1.15 1.1 1 g4 0
g8 0.87 1.32 1.35 1.13
g9 …
-0.33 1.01 1.38 1.21
g 10 0.10 0.85 1.03 1
… Distance Matrix
Original Data Matrix

18
 Types of Attributes

• Discrete
– Has only a finite or countably infinite set of values
– Examples: zip codes, counts, or the set of words in a collection of
documents
– Note: binary attributes are a special case of discrete attributes
• Ordinal
– Has only a finite or countably infinite set of values
– Order of values is important
– Examples: rankings (e.g., pain level 1-10), grades (A, B, C, D)
• Continuous
– Has real numbers as attribute values
– Examples: temperature, height, or weight
– Continuous attributes are typically represented as floating-point
variables

19
 Similarity/Dissimilarity for Simple Attributes

p and q are the attribute values for two data objects.

Discrete

Ordinal

Continuous

Dissimilarity and similarity between p and q

20
 Minkowski Distance—Continuous Attribute
• Minkowski distance: a generalization
d (i, j)  q | x  x |q  | x  x |q ... | x  x |q (q  0)
i1 j1 i2 j2 ip jp

• If q = 2, d is Euclidean distance
• If q = 1, d is Manhattan distance

Xi (1,7)
xi
12
8.48
q=2 6 q=1

6
Xj(7,1) xj
21
 Standardization

• Calculate the mean absolute deviation

mf  1
n (x1 f  x2 f  ...  xnf )
.

sf  1
n (| x1 f  m f |  | x2 f  m f | ... | xnf  m f |)

• Calculate the standardized measurement (z-score)

xif  m f
zif  sf

22
 Mahalanobis Distance
1
d ( p, q)  ( p  q)  ( p  q) T

 is the covariance matrix of the

input data X

1 n
 j ,k   ( X ij  X j )( X ik  X k )
n  1 i 1

Belongs to the family of bregman

divergence

For red points, the Euclidean distance is 14.7, Mahalanobis distance is 6.

23
 Mahalanobis Distance
Covariance Matrix:

0.3 0.2
 
 0.2 0.3
C

B A: (0.5, 0.5)
B: (0, 1)
A C: (1.5, 1.5)

Mahal(A,B) = 5
Mahal(A,C) = 4

24
 Common Properties of a Distance

• Distances, such as the Euclidean distance, have

some well known properties
1. d(p, q)  0 for all p and q and d(p, q) = 0 only if
p = q. (Positive definiteness)
2. d(p, q) = d(q, p) for all p and q. (Symmetry)
3. d(p, r)  d(p, q) + d(q, r) for all points p, q, and r.
(Triangle Inequality)
where d(p, q) is the distance (dissimilarity) between points
(data objects), p and q.

• A distance that satisfies these properties is a

metric

25
 Similarity for Binary Attributes
• Common situation is that objects, p and q, have only
binary attributes
• Compute similarities using the following quantities
M01 = the number of attributes where p was 0 and q was 1
M10 = the number of attributes where p was 1 and q was 0
M00 = the number of attributes where p was 0 and q was 0
M11 = the number of attributes where p was 1 and q was 1

• Simple Matching and Jaccard Coefficients

SMC = number of matches / total number of attributes
= (M11 + M00) / (M01 + M10 + M11 + M00)

J = number of matches / number of not-both-zero attributes values

= (M11) / (M01 + M10 + M11)

26
 SMC versus Jaccard: Example

p= 1000000000
q= 0000001001

M01 = 2 (the number of attributes where p was 0 and q was 1)

M10 = 1 (the number of attributes where p was 1 and q was 0)
M00 = 7 (the number of attributes where p was 0 and q was 0)
M11 = 0 (the number of attributes where p was 1 and q was 1)

SMC = (M11 + M00)/(M01 + M10 + M11 + M00) = (0+7) / (2+1+0+7) = 0.7

J = (M11) / (M01 + M10 + M11) = 0 / (2 + 1 + 0) = 0

27
 Document Data
• Each document becomes a `term' vector,
– each term is a component (attribute) of the vector,
– the value of each component is the number of times the
corresponding term occurs in the document.

timeout

season
coach

game
score
team

ball

lost
pla

wi
n
y

Document 1 3 0 5 0 2 6 0 2 0 2

Document 2 0 7 0 2 1 0 0 3 0 0

Document 3 0 1 0 0 1 2 2 0 3 0

28
 Cosine Similarity
• If d1 and d2 are two document vectors, then
cos( d1, d2 ) = (d1  d2) / ||d1|| ||d2|| ,
where  indicates vector dot product and || d || is the length of vector d.

• Example:
d1 = 3 2 0 5 0 0 0 2 0 0
d2 = 1 0 0 0 0 0 0 1 0 2

d1  d2= 3*1 + 2*0 + 0*0 + 5*0 + 0*0 + 0*0 + 0*0 + 2*1 + 0*0 + 0*2 = 5
||d1|| = (3*3+2*2+0*0+5*5+0*0+0*0+0*0+2*2+0*0+0*0)0.5 = (42) 0.5 = 6.481
||d2|| = (1*1+0*0+0*0+0*0+0*0+0*0+0*0+1*1+0*0+2*2) 0.5 = (6) 0.5 = 2.245

cos( d1, d2 ) = .3150

29
 Correlation

• Correlation measures the linear relationship between objects

• To compute correlation, we standardize data objects, p and q,
and then take their dot product (continuous attributes)

pk  ( pk  mean( p)) / std ( p)

qk  (qk  mean(q)) / std (q)

s( p, q)  p  q
30
 Common Properties of a Similarity

• Similarities, also have some well known

properties.
1. s(p, q) = 1 (or maximum similarity) only if p = q.

2. s(p, q) = s(q, p) for all p and q. (Symmetry)

where s(p, q) is the similarity between points (data

objects), p and q.

31
 Characteristics of the Input Data Are Important

• Sparseness
• Attribute type
• Type of Data
• Dimensionality
• Noise and Outliers
• Type of Distribution
• => Conduct preprocessing and select the appropriate
dissimilarity or similarity measure
• => Determine the objective of clustering and choose
the appropriate method

32
 Clustering Basics

• Definition and Motivation

• Data Preprocessing and Distance computation
• Objective of Clustering
• Clustering Evaluation

33
 Considerations for Cluster Analysis
• Partitioning criteria
– Single level vs. hierarchical partitioning (often, multi-level hierarchical
partitioning is desirable)
• Separation of clusters
– Exclusive (e.g., one customer belongs to only one region) vs. overlapping
(e.g., one document may belong to more than one topic)

• Hard versus fuzzy

– In fuzzy clustering, a point belongs to every cluster with some weight
between 0 and 1
– Weights must sum to 1
– Probabilistic clustering has similar characteristics

• Similarity measure and data types

• Heterogeneous versus homogeneous
– Cluster of widely different sizes, shapes, and densities
34
 Requirements of Clustering

• Scalability
• Ability to deal with different types of attributes
• Minimal requirements for domain knowledge to determine
input parameters
• Able to deal with noise and outliers
• Discovery of clusters with arbitrary shape
• Insensitive to order of input records
• High dimensionality
• Incorporation of user-specified constraints
• Interpretability and usability
• What clustering results we want to get?
35
Notion of a Cluster can be Ambiguous

How many clusters? Six Clusters

Two Clusters Four Clusters

36
 Partitional Clustering

Input Data A Partitional Clustering

37
 Hierarchical Clustering

p1
p3 p4
p2
p1 p2 p3 p4

Clustering Solution 1

p1
p3 p4
p2
p1 p2 p3 p4

Clustering Solution 2
38
 Types of Clusters: Center-Based

• Center-based
– A cluster is a set of objects such that an object in a cluster is closer
(more similar) to the “center” of a cluster, than to the center of any
other cluster
– The center of a cluster is often a centroid, the average of all the
points in the cluster, or a medoid, the most “representative” point
of a cluster

4 center-based clusters

39
 Types of Clusters: Density-Based

• Density-based
– A cluster is a dense region of points, which is separated by low-
density regions, from other regions of high density.
– Used when the clusters are irregular or intertwined, and when noise
and outliers are present.

6 density-based clusters

40
 Clustering Basics

• Definition and Motivation

• Data Preprocessing and Distance computation
• Objective of Clustering
• Clustering Evaluation

41
 Cluster Validation

• Cluster validation
– Quality: “goodness” of clusters
– Assess the quality and reliability of clustering
results

• Why validation?
– To avoid finding clusters formed by chance
– To compare clustering algorithms
– To choose clustering parameters
• e.g., the number of clusters

42
 Aspects of Cluster Validation

• Comparing the clustering results to ground truth

(externally known results)
– External Index
• Evaluating the quality of clusters without reference
to external information
– Use only the data
– Internal Index
• Determining the reliability of clusters
– To what confidence level, the clusters are not formed
by chance
– Statistical framework

43
 Comparing to Ground Truth

• Notation
– N: number of objects in the data set
– P={P1,…,Ps}: the set of “ground truth” clusters
– C={C1,…,Ct}: the set of clusters reported by a clustering
algorithm
• The “incidence matrix”
– N  N (both rows and columns correspond to objects)
– Pij = 1 if Oi and Oj belong to the same “ground truth” cluster
in P; Pij=0 otherwise
– Cij = 1 if Oi and Oj belong to the same cluster in C; Cij=0
otherwise

44
 Rand Index and Jaccard Coefficient

• A pair of data object (Oi,Oj) falls into one of the

following categories
– SS: Cij=1 and Pij=1; (agree)
– DD: Cij=0 and Pij=0; (agree)
– SD: Cij=1 and Pij=0; (disagree)
– DS: Cij=0 and Pij=1; (disagree)

| SS |  | DD |
• Rand index Rand 
| Agree |

| Agree |  | Disagree | | SS |  | SD |  | DS |  | DD |

– may be dominated by DD
• Jaccard Coefficient Jaccard coefficien t 
| SS |
| SS |  | SD |  | DS |

45
Clustering

g1 g2 g3 g4 g5

g1 1 1 1 0 0

g2 1 1 1 0 0 Clustering
g3 1 1 1 0 0 Same Different
Cluster Cluster
g4 0 0 0 1 1 Ground Same
truth 9 4
Cluster
g5 0 0 0 1 1 Different
4 8
Cluster
Groundtruth

g1 g2 g3 g4 g5
| SS |  | DD | 17
g1 1 1 0 0 0 Rand  
| SS |  | SD |  | DS |  | DD | 25
g2 1 1 0 0 0

g3 0 0 1 1 1 | SS | 9
Jaccard  
g4 0 0 1 1 1 | SS |  | SD |  | DS | 17
g5 0 0 1 1 1
46
 Entropy and Purity

• Notation
– | Ck  Pj | the number of objects in both the k-th cluster of
the clustering solution and j-th cluster of the groundtruth
– | Ck | the number of objects in the k-th cluster of the
clustering solution
– | Pj | the number of objects in the j-th cluster of the
groundtruth
1
• Purity Purity  
N k
max j | Ck  Pj |

• Normalized Mutual Information

| Ck  Pj | N  | Ck  Pj |
I (C , P) I (C , P)   log
NMI  k j N | Ck || Pj |
H (C ) H ( P)
| Ck | |C |
H (C )  
| Pj | | Pj |
log k H ( P)   log
k N N j N N 47
 Example
P1 P2 P3 P4 P5 P6 Total

C1 3 5 40 506 96 27 677

1
C2 4 7 280 29 39 2 361
Purity   max j | Ck  Pj |
N k
C3 1 1 1 7 4 671 685
506  280  671  162  331  358
C4 10 162 3 119 73 2 369 Purity 
3204
C5 331 22 5 70 13 23 464  0.7203

C6 5 358 12 212 48 13 648

total 354 555 341 943 273 738 3204

| Ck  Pj | N  | Ck  Pj |
NMI 
I (C , P) I (C , P)   log
H (C ) H ( P) k j N | Ck || Pj |
| Ck | |C |
H (C )  
| Pj | | Pj |
log k H ( P)   log
k N N j N N 48
 Internal Index

• “Ground truth” may be unavailable

• Use only the data to measure cluster quality
– Measure the “cohesion” and “separation” of clusters
– Calculate the correlation between clustering results
and distance matrix

49
 Cohesion and Separation
• Cohesion is measured by the within cluster sum of squares
WSS    ( x  mi ) 2
i xCi

• Separation is measured by the between cluster sum of squares

BSS   Ci (m  mi )2
i
where |Ci| is the size of cluster i, m is the centroid of the whole data set

• BSS + WSS = constant

• WSS (Cohesion) measure is called Sum of Squared Error (SSE)—a
commonly used measure
• A larger number of clusters tend to result in smaller SSE

50
 Example

m
  
1 m1 2 3 4 m2 5

WSS (1  3)2  (2  3)2  (4  3)2  (5  3)2  10

K=1 :
BSS  4  (3  3)2  0
Total  10  0  10
WSS (1  1.5) 2  (2  1.5) 2  (4  4.5) 2  (5  4.5) 2  1
K=2 :
BSS  2  (3  1.5) 2  2  (4.5  3) 2  9
Total  1  9  10

K=4: WSS (1  1) 2  (2  2) 2  (4  4) 2  (5  5) 2  0
BSS  1 (1  3) 2  1 (2  3) 2  1 (4  3) 2  1 (5  3) 2  10
Total  0  10  10

51
 Silhouette Coefficient
• Silhouette Coefficient combines ideas of both cohesion and separation

• For an individual point, i

– Calculate a = average distance of i to the points in its cluster
– Calculate b = min (average distance of i to points in another cluster)
– The silhouette coefficient for a point is then given by

s = 1 – a/b if a < b, (s = b/a - 1 if a  b, not the usual case)

– Typically between 0 and 1 b

– The closer to 1 the better a

• Can calculate the Average Silhouette width for a cluster or a clustering

52
 Correlation with Distance Matrix

• Distance Matrix
– Dij is the similarity between object Oi and Oj
• Incidence Matrix
– Cij=1 if Oi and Oj belong to the same cluster, Cij=0
otherwise
• Compute the correlation between the two
matrices
– Only n(n-1)/2 entries needs to be calculated
• High correlation indicates good clustering

53
 Correlation with Distance Matrix

• Given Distance Matrix D = {d11,d12, …, dnn } and Incidence

Matrix C= { c11, c12,…, cnn } .

• Correlation r between D and C is given by

n _ _

 (d
i 1, j 1
ij  d )(cij  c)
r
n _ n _

 (dij  d )
i 1, j 1
2
 ij
( c  c
i 1, j 1
) 2

54
 Are There Clusters in the Data?
1 1

0.9 0.9

0.8 0.8

0.7 0.7

Random 0.6 0.6 DBSCAN

Points 0.5 0.5
y

y
0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x x
1 1

0.9 0.9

K-means 0.8 0.8

Complete
0.7 0.7
Link
0.6 0.6

0.5 0.5
y

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x x
55
Measuring Cluster Validity Via Correlation

• Correlation of incidence and distance matrices for the K-

means clusterings of the following two data sets

1 1

0.9 0.9

0.8 0.8

0.7 0.7

0.6 0.6

0.5 0.5
y

y
0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x x

Corr = -0.9235 Corr = -0.5810

56
Using Similarity Matrix for Cluster Validation

• Order the similarity matrix with respect to cluster

labels and inspect visually.
1
1
10 0.9
0.9
20 0.8
0.8
30 0.7
0.7
40 0.6
0.6

Points
50 0.5
0.5
y

60 0.4
0.4
70 0.3
0.3
80 0.2
0.2
90 0.1
0.1
100 0
0 20 40 60 80 100 Similarity
0 0.2 0.4 0.6 0.8 1
Points
x

57
Using Similarity Matrix for Cluster Validation

• Clusters in random data are not so crisp

1
1
10 0.9
0.9
20 0.8
0.8
30 0.7
0.7
40 0.6
0.6

Points
50 0.5
0.5
y

60 0.4
0.4
70 0.3
0.3
80 0.2
0.2
90 0.1
0.1
100 0
0 20 40 60 80 100 Similarity
0 0.2 0.4 0.6 0.8 1
Points
x

58
 Reliability of Clusters

• Need a framework to interpret any measure

– For example, if our measure of evaluation has the

value, 10, is that good, fair, or poor?

• Statistics provide a framework for cluster validity

– The more “atypical” a clustering result is, the more
likely it represents valid structure in the data

59
 Statistical Framework for SSE
• Example
– Compare SSE of 0.005 against three clusters in random data
– SSE Histogram of 500 sets of random data points of size 100—
lowest SSE is 0.0173
1
50
0.9
45
0.8
40
0.7
35
0.6
30

Count
0.5
y

25
0.4
20
0.3
15
0.2
10
0.1
5
0
0 0.2 0.4 0.6 0.8 1 0
0.016 0.018 0.02 0.022 0.024 0.026 0.028 0.03 0.032 0.034
x SSE

SSE = 0.005
60
Determine the Number of Clusters Using SSE

• SSE curve
10

6 9

8
4
7

2 6

SSE
5
0
4
-2 3

2
-4
1
-6 0
2 5 10 15 20 25 30
5 10 15
K

Clustering of Input Data SSE wrt K

61
 Take-away Message

• What’s clustering?
• Why clustering is important?
• How to preprocess data and compute
dissimilarity/similarity from data?
• What’s a good clustering solution?
• How to evaluate the clustering results?

62