DATA MINING

LECTURE 8
Clustering Validation
Minimum Description Length
Information Theory
Co-Clustering
CLUSTERING VALIDITY
Cluster Validity
• How do we evaluate the “goodness” of the resulting
clusters?

• But “clustering lies in the eye of the beholder”!

• Then why do we want to evaluate them?

• To avoid finding patterns in noise
• To compare clustering algorithms
• To compare two clusterings
• To compare two clusters
 Clusters found in Random 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 Complete
0.8 0.8

0.7 0.7

0.6 0.6
Link
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
Different Aspects of Cluster Validation
1. Determining the clustering tendency of a set of data, i.e.,
distinguishing whether non-random structure actually exists in the
data.
2. Comparing the results of a cluster analysis to externally known
results, e.g., to externally given class labels.
3. Evaluating how well the results of a cluster analysis fit the data
without reference to external information.
- Use only the data
4. Comparing the results of two different sets of cluster analyses to
determine which is better.
5. Determining the ‘correct’ number of clusters.

For 2, 3, and 4, we can further distinguish whether we want to

evaluate the entire clustering or just individual clusters.
Measures of Cluster Validity
• Numerical measures that are applied to judge various aspects
of cluster validity, are classified into the following three types.
• External Index: Used to measure the extent to which cluster labels
match externally supplied class labels.
• E.g., entropy, precision, recall
• Internal Index: Used to measure the goodness of a clustering
structure without reference to external information.
• E.g., Sum of Squared Error (SSE)
• Relative Index: Used to compare two different clusterings or
clusters.
• Often an external or internal index is used for this function, e.g., SSE or
entropy
• Sometimes these are referred to as criteria instead of indices
• However, sometimes criterion is the general strategy and index is the
numerical measure that implements the criterion.
 Measuring Cluster Validity Via Correlation
• Two matrices
• Similarity or Distance Matrix
• One row and one column for each data point
• An entry is the similarity or distance of the associated pair of points
• “Incidence” Matrix
• One row and one column for each data point
• An entry is 1 if the associated pair of points belong to the same cluster
• An entry is 0 if the associated pair of points belongs to different clusters
• Compute the correlation between the two matrices
• Since the matrices are symmetric, only the correlation between
n(n-1) / 2 entries needs to be calculated.
• High correlation (positive for similarity, negative for
distance) indicates that points that belong to the same
cluster are close to each other.
• Not a good measure for some density or contiguity based
clusters.
Measuring Cluster Validity Via Correlation
• Correlation of incidence and proximity 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

y
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

Corr = -0.9235 Corr = -0.5810

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

Points
0.6 50 0.5

0.5 60 0.4
y

0.4 70 0.3

0.3 80 0.2

0.2 90 0.1

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

  𝑑 𝑖𝑗 − 𝑑 𝑚𝑖𝑛
𝑠 𝑖𝑚(𝑖 , 𝑗 )=1 −
𝑑 𝑚𝑎𝑥 −𝑑 𝑚𝑖𝑛
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

DBSCAN
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

K-means
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

Complete Link
Using Similarity Matrix for Cluster Validation
1

0.9
1 500
0.8
2 6
0.7
1000
3 0.6
4
1500 0.5

0.4
2000
0.3
5
0.2
2500
0.1
7
3000 0
500 1000 1500 2000 2500 3000

DBSCAN
Internal Measures: SSE
• Internal Index: Used to measure the goodness of a
clustering structure without reference to external
information
• Example: SSE
• SSE is good for comparing two clusterings or two clusters
(average SSE).
• Can also be used to estimate the number of clusters
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
Estimating the “right” number of clusters
• Typical approach: find a “knee” in an internal measure curve.
10

6
SSE

0
2 5 10 15 20 25 30
K

• Question: why not the k that minimizes the SSE?

• Forward reference: minimize a measure, but with a “simple” clustering
• Desirable property: the clustering algorithm does not require the
number of clusters to be specified (e.g., DBSCAN)
Internal Measures: SSE
• SSE curve for a more complicated data set

1
2 6

3
4

SSE of clusters found using K-means

Internal Measures: Cohesion and Separation
• Cluster Cohesion: Measures how closely related
are objects in a cluster
• Cluster Separation: Measure how distinct or well-
separated a cluster is from other clusters
• Example: Squared Error
• Cohesion is measured by the within cluster sum of squares (SSE)

WSS    ( x  ci ) 2
i xCi

• Separation is measured by the between cluster sum of squares

BSS   m (c  c )
i i
2

• Where mi is the size of cluster i

 Internal Measures: Cohesion and Separation
• A proximity graph based approach can also be used for
cohesion and separation.
• Cluster cohesion is the sum of the weight of all links within a cluster.
• Cluster separation is the sum of the weights between nodes in the cluster
and nodes outside the cluster.

cohesion separation
Internal measures – caveats
• Internal measures have the problem that the
clustering algorithm did not set out to optimize
this measure, so it is will not necessarily do well
with respect to the measure.

• An internal measure can also be used as an

objective function for clustering
 Framework for Cluster Validity
• 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 “non-random” a clustering result is, the more likely it
represents valid structure in the data
• Can compare the values of an index that result from random data or
clusterings to those of a clustering result.
• If the value of the index is unlikely, then the cluster results are valid
• For comparing the results of two different sets of cluster
analyses, a framework is less necessary.
• However, there is the question of whether the difference between two
index values is significant
Statistical Framework for SSE
• Example
• Compare SSE of 0.005 against three clusters in random data
• Histogram of SSE for three clusters in 500 random data sets of
100 random points distributed in the range 0.2 – 0.8 for x and y
• Value 0.005 is very unlikely
1
50
0.9
45
0.8
40
0.7
35
0.6
Count 30
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
Statistical Framework for Correlation
• Correlation of incidence and proximity 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

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

Empirical p-value
• If we have a measurement v (e.g., the SSE value)
• ..and we have N measurements on random datasets
• …the empirical p-value is the fraction of measurements in
the random data that have value less or equal than value
v (or greater or equal if we want to maximize)
• i.e., the value in the random dataset is at least as good as that in
the real data

• We usually require that p-value ≤ 0.05

• Hard question: what is the right notion of a random

dataset?
External Measures for Clustering Validity
• Assume that the data is labeled with some class labels
• E.g., documents are classified into topics, people classified
according to their income, senators classified as republican or
democrat.
• In this case we want the clusters to be homogeneous
with respect to classes
• Each cluster should contain elements of mostly one class
• Also each class should ideally be assigned to a single cluster
• This does not always make sense
• Clustering is not the same as classification
• But this is what people use most of the time
 Measures
•• = number of points
 
• = points in cluster i
Class 1 Class 2 Class 3
• = points in class j
• = points in cluster i coming from Cluster 1
class j
Cluster 2
• = prob of element from class j in
cluster i Cluster 3
• Entropy:
• Of a cluster i:
• Highest when uniform, zero when single
class
• Of a clustering:
• Purity:
• Of a cluster i:
• Of a clustering:
Measures
•  Precision:
• Of cluster i with respect to class j:
• For the precision of a clustering you can take the maximum

• Recall:
• Of cluster i with respect to class j:
• For the precision of a clustering you can take the maximum

• F-measure:
• Harmonic Mean of Precision and Recall:
 Good and bad clustering

Class 1 Class 2 Class 3 Class 1 Class 2 Class 3

Cluster 1 Cluster 1

Cluster 2 Cluster 2

Cluster 3 Cluster 3

100 300 100 300

Purity: (0.94, 0.81, 0.85) – overall 0.86 Purity: (0.38, 0.38, 0.38) – overall 0.38
Precision: (0.94, 0.81, 0.85) Precision: (0.38, 0.38, 0.38)
Recall: (0.85, 0.9, 0.85) Recall: (0.35, 0.42, 0.38)
Another bad clustering

Class 1 Class 2 Class 3

Cluster 1

Cluster 2 Cluster 1:
Purity: 1
Cluster 3 Precision: 1
Recall: 0.35
100 300
External Measures of Cluster Validity:
Entropy and Purity
Final Comment on Cluster Validity
“The validation of clustering structures is the most
difficult and frustrating part of cluster analysis.
Without a strong effort in this direction, cluster
analysis will remain a black art accessible only to
those true believers who have experience and
great courage.”

Algorithms for Clustering Data, Jain and Dubes

MINIMUM DESCRIPTION
LENGTH
Occam’s razor
• Most data mining tasks can be described as
creating a model for the data
• E.g., the EM algorithm models the data as a mixture of
Gaussians, the K-means models the data as a set of
centroids.
• What is the right model?

• Occam’s razor: All other things being equal, the

simplest model is the best.
• A good principle for life as well
Occam's Razor and MDL
• What is a simple model?

• Minimum Description Length Principle: Every model

provides a (lossless) encoding of our data. The model
that gives the shortest encoding (best compression)
of the data is the best.
• Related: Kolmogorov complexity. Find the shortest program
that produces the data (uncomputable).
• MDL restricts the family of models considered

• Encoding cost: cost of party A to transmit to party B the data.

Minimum Description Length (MDL)
• The description length consists of two terms
• The cost of describing the model (model cost)
• The cost of describing the data given the model (data cost).
• L(D) = L(M) + L(D|M)

• There is a tradeoff between the two costs

• Very complex models describe the data in a lot of detail but are
expensive to describe the model
• Very simple models are cheap to describe but it is expensive to
describe the data given the model

• This is generic idea for finding the right model

• We use MDL as a blanket name.
 35

Example
• Regression: find a polynomial for describing a set of values
• Model complexity (model cost): polynomial coefficients
• Goodness of fit (data cost): difference between real value and the polynomial value

Minimum model cost High model cost Low model cost

High data cost Minimum data cost Low data cost

MDL avoids overfitting automatically!

Source: Grunwald et al. (2005) Tutorial on MDL.
Example
• Suppose you want to describe a set of integer numbers
• Cost of describing a single number is proportional to the value of the number x (e.g.,
logx).
• How can we get an efficient description?

• Cluster integers into two clusters and describe the cluster by the centroid
and the points by their distance from the centroid
• Model cost: cost of the centroids
• Data cost: cost of cluster membership and distance from centroid

• What are the two extreme cases?

MDL and Data Mining
• Why does the shorter encoding make sense?
• Shorter encoding implies regularities in the data
• Regularities in the data imply patterns
• Patterns are interesting

• Example

00001000010000100001000010000100001000010001000010000100001

• Short description length, just repeat 12 times 00001

0100111001010011011010100001110101111011011010101110010011100

• Random sequence, no patterns, no compression

Is everything about compression?
• Jürgen Schmidhuber:
A theory about creativity, art and fun
• Interesting Art corresponds to a novel pattern that we cannot
compress well, yet it is not too random so we can learn it
• Good Humor corresponds to an input that does not compress
well because it is out of place and surprising
• Scientific discovery corresponds to a significant compression
event
• E.g., a law that can explain all falling apples.

• Fun lecture:
• Compression Progress: The Algorithmic Principle Behind Curio
sity and
Creativity
Issues with MDL
• What is the right model family?
• This determines the kind of solutions that we can have
• E.g., polynomials
• Clusterings

• What is the encoding cost?

• Determines the function that we optimize
• Information theory
INFORMATION THEORY
A short introduction
Encoding
• Consider the following sequence

AAABBBAAACCCABACAABBAACCABAC

• Suppose you wanted to encode it in binary form,

how would you do it?
50% A A is 50% of the sequence A0
25% B We should give it a shorter B  10
representation
25% C C  11
This is actually provably the best encoding!
Encoding
•• Prefix
  Codes: no codeword is a prefix of another
A0 Uniquely directly decodable
B  10 For every code we can find a prefix code
C  11 of equal length

• Codes and Distributions: There is one to one mapping

between codes and distributions
• If P is a distribution over a set of elements (e.g., {A,B,C}) then there
exists a (prefix) code C where
• For every (prefix) code C of elements {A,B,C}, we can define a
distribution

• The code defined has the smallest average codelength!

Entropy
•• Suppose
  we have a random variable X that takes n distinct values

that have probabilities

• This defines a code C with . The average codelength is

• This (more or less) is the entropy of the random variable X

• Shannon’s theorem: The entropy is a lower bound on the average

codelength of any code that encodes the distribution P(X)
• When encoding N numbers drawn from P(X), the best encoding length we
can hope for is
• Reminder: Lossless encoding
Entropy
• 
• What does it mean?
• Entropy captures different aspects of a distribution:
• The compressibility of the data represented by random
variable X
• Follows from Shannon’s theorem
• The uncertainty of the distribution (highest entropy for
uniform distribution)
• How well can I predict a value of the random variable?
• The information content of the random variable X
• The number of bits used for representing a value is the information
content of this value.
Claude Shannon
Father of Information Theory

Envisioned the idea of communication

of information with 0/1 bits

Introduced the word “bit”

The word entropy was suggested by Von Neumann

• Similarity to physics, but also
• “nobody really knows what entropy really is, so in any
conversation you will have an advantage”
Some information theoretic measures
•  Conditional entropy H(Y|X): the uncertainty for Y
given that we know X

• Mutual Information I(X,Y): The reduction in the

uncertainty for Y (or X) given that we know X (or
Y)
Some information theoretic measures
•  Cross Entropy: The cost of encoding distribution
P, using the code of distribution Q

• KL Divergence KL(P||Q): The increase in

encoding cost for distribution P when using the
code of distribution Q

• Not symmetric
• Problematic if Q not defined for all x of P.
Some information theoretic measures
•  Jensen-Shannon Divergence JS(P,Q): distance
between two distributions P and Q
• Deals with the shortcomings of KL-divergence

• If M = ½ (P+Q) is the mean distribution

• Jensen-Shannon is a metric
USING MDL FOR
CO-CLUSTERING
(CROSS-ASSOCIATIONS)
Thanks to Spiros Papadimitriou.
Co-clustering
• Simultaneous grouping of rows and columns of a
matrix into homogeneous groups
Students buying books

5
5

97% 3%

Customer groups
10
10

54% 15
Customers

15

20
20 3% 96%

25
25 5 10 15 20 25
5
Products
10 15 20 25
Product groups

CEOs buying BMWs

Co-clustering

• Step 1: How to define a “good” partitioning?

Intuition and formalization

• Step 2: How to find it?

 Co-clustering
Intuition

Why is this
better?

Row groups
Row groups

versus

Column groups Column groups

1. Similar nodes are

grouped together A few,
Good Good
homogeneous
Clustering 2. As few groups as Compression
blocks
necessary

implies
 Co-clustering
MDL formalization—Cost objective
ℓ = 3 col. groups
m1 m2 m3
density of ones

n1 n1m2 H(p1,2) bits for (1,2)

p1,1 p1,2 p1,3
block size entropy
k = 3 row groups

 i,j nimj H(pi,j) bits total

n2 p2,1 p2,2 p2,3 data cost
+
model cost
n3 p3,1 p3,2 p3,3 +
row-partition col-partition
description description

n × m matrix
+ log*k + log*ℓ +
transmit
 log n m 
i,j
i
transmit
j

#partitions #ones ei,j

Co-clustering
MDL formalization—Cost objective


one row group n row groups 
one col group m col groups

high code cost low

(block contents)

+
low high
description cost
(block structure)
Co-clustering
MDL formalization—Cost objective


k = 3 row groups
ℓ = 3 col groups

low code cost

(block contents)

low
+
description cost
(block structure)
 one row group n row groups k = 3 row groups
one col group m col groups ℓ = 3 col groups
MDL formalization—Cost objective

ℓ
Cost vs. number of groups
Co-clustering

k
total bit cost
Co-clustering

• Step 1: How to define a “good” partitioning?

Intuition and formalization

• Step 2: How to find it?

 Search for solution
Overview: assignments w/ fixed number of groups (shuffles)

original groups row shuffle column shuffle row shuffle

reassign all rows, reassign all columns,

No cost improvement:
holding column holding row Discard
assignments fixed assignments fixed
Search for solution
Overview: assignments w/ fixed number of groups (shuffles)
Final shuffle result
row shuffle column shuffle column shuffle

column
row shuffle
shuffle

No cost improvement:
Discard
Search for solution
Shuffles
• Let
Similarity (“KL-divergences”)
p1,1 p1,2 p1,3 of row fragments
denote row and col. partitions
to blocks of aat
rowthe I-th iteration
group
• Fix I and for everyAssign
row x:to second row-group
p2,1 p p
2,2 2,3
• Splice into ℓ parts, one for each column group
• Let j, for j = 1,…,ℓ, be the number of ones in each part
p3,1 p3,2
• Assign row x to pthe
3,3
row group i¤  I+1(x) such that, for all
i = 1,…,k,
Search for solution
Overview: number of groups k and ℓ (splits & shuffles)
k = 5, ℓ = 5
 Search for solution
Overview: number of groups k and ℓ (splits & shuffles)
k = 1, ℓ = 1

shuffle shuffle
col. split
row row split

No cost improvement:
Discard
k = 5,
6, ℓ = 56 k = 5, ℓ = 5
shuffle shuffle shuffle shuffle shuffle shuffle shuffle
col. splitrow split col. split row split col. split row split col. split

k=1, ℓ=2 k=2, ℓ=2 k=2, ℓ=3 k=3, ℓ=3 k=3, ℓ=4 k=4, ℓ=4 k=4, ℓ=5

Split: Shuffle:
Increase k or ℓ Rearrange rows or cols
 Search for solution
Overview: number of groups k and ℓ (splits & shuffles)
k = 1, ℓ = 1 k = 5, ℓ = 5

k = 5, ℓ = 5

Final result

k=1, ℓ=2 k=2, ℓ=2 k=2, ℓ=3 k=3, ℓ=3 k=3, ℓ=4 k=4, ℓ=4 k=4, ℓ=5

Split: Shuffle:
Increase k or ℓ Rearrange rows or cols
 Co-clustering
CLASSIC
CLASSIC corpus
• 3,893 documents
• 4,303 words
Documents

• 176,347 “dots” (edges)

Combination of 3 sources:
Words
• MEDLINE (medical)
• CISI (info. retrieval)
• CRANFIELD (aerodynamics)
 Graph co-clustering
CLASSIC
Documents

Words

“CLASSIC” graph of documents & words:

k = 15, ℓ = 19
 Co-clustering
CLASSIC
insipidus, alveolar, aortic, death, blood, disease, clinical, paint, examination, fall,
prognosis, intravenous cell, tissue, patient raise, leave, based

MEDLINE
(medical)

CISI
(Information
Retrieval)

CRANFIELD
(aerodynamics)

providing, studying, records, abstract, notation, works, shape, nasa, leading,

development, students, rules construct, bibliographies assumed, thin

“CLASSIC” graph of documents & words:

k = 15, ℓ = 19
Co-clustering
CLASSIC

Document Document class Precision

cluster # CRANFIELD CISI MEDLINE

1 0 1 390 0.997 0.999

2 0 0 610 1.000
3 2 676 9 0.984
0.975
4 1 317 6 0.978
5 3 452 16 0.960

0.94-1.00
6 207 0 0 1.000
7 188 0 0 1.000
8 131 0 0 1.000
9 209 0 0 1.000
10 107 2 0 0.982 0.987
11 152 3 2 0.968
12 74 0 0 1.000
13 139 9 0 0.939
14 163 0 0 1.000
15 24 0 0 1.000
Recall 0.996 0.990 0.968
0.97-0.99