Data Warehousing/Mining

Comp 150 DW
Chapter 8. Cluster Analysis

Instructor: Dan Hebert

Data Warehousing/Mining
 Chapter 8. Cluster Analysis
 What is Cluster Analysis?
 Types of Data in Cluster Analysis
 A Categorization of Major Clustering Methods
 Partitioning Methods
 Hierarchical Methods
 Density-Based Methods
 Grid-Based Methods
 Model-Based Clustering Methods
 Outlier Analysis
 Summary
Data Warehousing/Mining
 General Applications of Clustering

 Pattern Recognition
 Spatial Data Analysis
– create thematic maps in GIS by clustering feature spaces
– detect spatial clusters and explain them in spatial data mining
 Image Processing
 Economic Science (especially market research)
 WWW
– Document classification
– Cluster Weblog data to discover groups of similar access patterns

Data Warehousing/Mining
 Examples of Clustering Applications

 Marketing: Help marketers discover distinct groups in

their customer bases, and then use this knowledge to
develop targeted marketing programs
 Land use: Identification of areas of similar land use in an
earth observation database
 Insurance: Identifying groups of motor insurance policy
holders with a high average claim cost
 City-planning: Identifying groups of houses according to
their house type, value, and geographical location
 Earth-quake studies: Observed earth quake epicenters
should be clustered along continent faults
Data Warehousing/Mining
 What Is Good Clustering?

 A good clustering method will produce high quality

clusters with
– high intra-class similarity
– low inter-class similarity
 The quality of a clustering result depends on both the
similarity measure used by the method and its
implementation.
 The quality of a clustering method is also measured by its
ability to discover some or all of the hidden patterns.

Data Warehousing/Mining
 Requirements of Clustering in Data
Mining
 Scalability
 Ability to deal with different types of attributes
 Discovery of clusters with arbitrary shape
 Minimal requirements for domain knowledge to
determine input parameters
 Able to deal with noise and outliers
 Insensitive to order of input records
 High dimensionality
 Incorporation of user-specified constraints
 Interpretability and usability
Data Warehousing/Mining
 Measure the Quality of Clustering

 Dissimilarity/Similarity metric: Similarity is expressed in

terms of a distance function, which is typically metric:
d(i, j)
 There is a separate “quality” function that measures the
“goodness” of a cluster.
 The definitions of distance functions are usually very
different for interval-scaled, boolean, categorical, ordinal
and ratio variables.
 Weights should be associated with different variables
based on applications and data semantics.
 It is hard to define “similar enough” or “good enough”
– the answer is typically highly subjective.

Data Warehousing/Mining
 Type of data in clustering analysis
 Interval-scaled variables:
 Binary variables:
 Nominal, ordinal, and ratio variables:
 Variables of mixed types:

Data Warehousing/Mining
 Interval-valued variables

 Standardize data
– Calculate the mean absolute deviation:
s f  1n (| x1 f  m f |  | x2 f  m f | ... | xnf  m f |)

where m f  1n (x1 f  x2 f  ...  xnf )

.

– Calculate the standardized measurement (z-score)

xif  m f
zif  sf
 Using mean absolute deviation is more robust than using
standard deviation

Data Warehousing/Mining 1
 Similarity and Dissimilarity Between
Objects
 Distances are normally used to measure the
similarity or dissimilarity between two data
objects
 Some popular ones include: Minkowski distance:
d (i, j)  (| x  x | | x  x | ... | x  x | )
q q q 1/q
i1 j1 i2 j2 ip jp
where i = (xi1, xi2, …, xip) and j = (xj1, xj2, …, xjp) are two
p-dimensional data objects, and q is a positive integer
 If q = 1, d is Manhattan distance
d (i, j) | x  x |  | x  x | ... | x  x |
i1 j1 i2 j 2 ip j p
Data Warehousing/Mining 1
 Similarity and Dissimilarity Between
Objects (Cont.)
 If q = 2, d is Euclidean distance:
d (i, j)  (| x  x | 2  | x  x |2 ... | x  x |2 )
i1 j1 i2 j2 ip jp

– Properties
 d(i,j)  0
 d(i,i) = 0
 d(i,j) = d(j,i)
 d(i,j)  d(i,k) + d(k,j)

Data Warehousing/Mining 1
 Binary Variables
 2 states – symmetric if no preference for which is 0 and 1
 A contingency table for binary data
Object j
1 0 sum a - i & j both equal 1
b - i=1, j=0
1 a b a b …
Object i 0 c d cd
sum a  c b  d p
 Simple matching coefficient (invariant, if the binary
d (i, j )  bc
variable is symmetric): a bc d
 Jaccard coefficient (noninvariant if the binary variable is
asymmetric)- ignore negative matches: d (i, j)  b  c
a bc
Data Warehousing/Mining 1
 Dissimilarity between Binary
Variables
 Example – do patients have the same disease
Name Gender Fever Cough Test-1 Test-2 Test-3 Test-4
Jack M Y N P N N N
Mary F Y N P N P N
Jim M Y P N N N N
– gender is a symmetric attribute (use Jaccard coefficeint so ignore )
– the remaining attributes are asymmetric binary
– let the values Y and P be set to 1, and the value N be set to 0
01
d ( jack , mary )   0.33 Most likely to have same disease
2 01
11
d ( jack , jim )   0.67
111
1 2
d ( jim , mary )   0.75 Unlikely to have same disease
11 2
Data Warehousing/Mining 1
 Nominal Variables

 A generalization of the binary variable in that it can take

more than 2 states, e.g., red, yellow, blue, green
 Method 1: Simple matching
– m: # of matches, p: total # of variables

d (i, j)  p 
p
m

 Method 2: use a large number of binary variables

– creating a new binary variable for each of the M nominal states

Data Warehousing/Mining 1
 Ordinal Variables

 An ordinal variable can be discrete or continuous

 order is important, e.g., rank (gold, silver, bronze)
 Can be treated like interval-scaled-use following steps
– f is variable from set of variables, value of f for ith object is xif
– replace each xif by its’ rank rif {1,..., M f }
– map the range of each variable onto [0, 1] by replacing i-th object
in the f-th variable by
rif 1
zif 
M f 1
– compute the dissimilarity using methods for interval-scaled
variables
Data Warehousing/Mining 1
 Ratio-Scaled Variables

 Ratio-scaled variable: a positive measurement on a

nonlinear scale, approximately at exponential scale,
such as AeBt or Ae-Bt (A & B are positive constants)
 Methods:
– treat them like interval-scaled variables — not a good choice! Scale
distorted
– apply logarithmic transformation

yif = log(xif)
– treat them as continuous ordinal data treat their rank as interval-
scaled.
Data Warehousing/Mining 1
 Variables of Mixed Types

 A database may contain all the six types of variables

– symmetric binary, asymmetric binary, nominal, ordinal, interval
and ratio.
 One may use a weighted formula to combine their
effects.  pf  1 ij( f ) d ij( f )
d (i, j ) 
 pf  1 ij( f )
– f is binary or nominal:
dij(f) = 0 if xif = xjf , or dij(f) = 1 o.w.
– f is interval-based: use the normalized distance
– f is ordinal or ratio-scaled
 compute ranks rif and
zif 
r
if 1
 and treat zif as interval-scaled M f 1

Data Warehousing/Mining 1
 Major Clustering Approaches
 Partitioning algorithms: Construct various partitions and then
evaluate them by some criterion
 Hierarchy algorithms: Create a hierarchical decomposition of
the set of data (or objects) using some criterion
 Density-based: based on connectivity and density functions
 Grid-based: based on a multiple-level granularity structure
 Model-based: A model is hypothesized for each of the clusters
and the idea is to find the best fit of that model to each other

Data Warehousing/Mining 1
 Partitioning Algorithms: Basic
Concept
 Partitioning method: Construct a partition of a database D
of n objects into a set of k clusters
 Given a k, find a partition of k clusters that optimizes the
chosen partitioning criterion
– Global optimal: exhaustively enumerate all partitions
– Heuristic methods: k-means and k-medoids algorithms
– k-means (MacQueen’67): Each cluster is represented by the center
of the cluster
– k-medoids or PAM (Partition around medoids) (Kaufman &
Rousseeuw’87): Each cluster is represented by one of the objects in
the cluster

Data Warehousing/Mining 2
 The K-Means Clustering Method
 Given k, the k-means algorithm is implemented
in 4 steps:
– Partition objects into k nonempty subsets
– Compute seed points as the centroids of the
clusters of the current partition. The centroid is the
center (mean point) of the cluster.
– Assign each object to the cluster with the nearest
seed point.
– Go back to Step 2, stop when no more new
assignment.

Data Warehousing/Mining 2
 The K-Means Clustering Method
 Example
10 10

9 9

8 8

7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

10 10

9 9

8 8

7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

Data Warehousing/Mining 2
 K-Means example

 2, 3, 6, 8, 9, 12, 15, 18, 22 – break into 3 clusters

– Cluster 1 - 2, 8, 15 – mean = 8.3
– Cluster 2 - 3, 9, 18 – mean = 10
– Cluster 3 - 6, 12, 22 – mean = 13.3
 Re-assign
– Cluster 1 - 2, 3, 6, 8, 9 – mean = 5.6
– Cluster 2 – mean = 0
– Cluster 3 – 12, 15, 18, 22 – mean = 16.75
 Re-assign
– Cluster 1 – 3, 6, 8, 9 – mean = 6.5
– Cluster 2 – 2 – mean = 2
– Cluster 3 = 12, 15, 18, 22 – mean = 16.75
Data Warehousing/Mining 2
 K-Means example (continued)

 Re-assign
– Cluster 1 - 6, 8, 9 – mean = 7.6
– Cluster 2 – 2, 3 – mean = 2.5
– Cluster 3 – 12, 15, 18, 22 – mean = 16.75
 Re-assign
– Cluster 1 - 6, 8, 9 – mean = 7.6
– Cluster 2 – 2, 3 - mean = 2.5
– Cluster 3 – 12, 15, 18, 22 – mean = 16.75
 No change, so we’re done

Data Warehousing/Mining 2
 K-Means example – different starting
order
 2, 3, 6, 8, 9, 12, 15, 18, 22 – break into 3 clusters
– Cluster 1 - 2, 12, 18 – mean = 10.6
– Cluster 2 - 6, 9, 22 – mean = 12.3
– Cluster 3 – 3, 8, 15 – mean = 8.6
 Re-assign
– Cluster 1 - mean = 0
– Cluster 2 – 12, 15, 18, 22 - mean = 16.75
– Cluster 3 – 2, 3, 6, 8, 9 – mean = 5.6
 Re-assign
– Cluster 1 – 2 – mean = 2
– Cluster 2 – 12, 15, 18, 22 – mean = 16.75
– Cluster 3 = 3, 6, 8, 9 – mean = 6.5
Data Warehousing/Mining 2
 K-Means example (continued)

 Re-assign
– Cluster 1 – 2, 3 – mean = 2.5
– Cluster 2 – 12, 15, 18, 22 – mean = 16.75
– Cluster 3 – 6, 8, 9 – mean = 7.6
 Re-assign
– Cluster 1 – 2, 3 – mean = 2.5
– Cluster 2 – 12, 15, 18, 22 - mean = 16.75
– Cluster 3 – 6, 8, 9 – mean = 7.6
 No change, so we’re done

Data Warehousing/Mining 2
 Comments on the K-Means Method

 Strength
– Relatively efficient: O(tkn), where n is # objects, k is #
clusters, and t is # iterations. Normally, k, t << n.
– Often terminates at a local optimum. The global optimum
may be found using techniques such as: deterministic
annealing and genetic algorithms
 Weakness
– Applicable only when mean is defined, then what about
categorical data?
– Need to specify k, the number of clusters, in advance
– Unable to handle noisy data and outliers
– Not suitable to discover clusters with non-convex shapes
Data Warehousing/Mining 2
 Variations of the K-Means Method
 A few variants of the k-means which differ in
– Selection of the initial k means
– Dissimilarity calculations
– Strategies to calculate cluster means
 Handling categorical data: k-modes (Huang’98)
– Replacing means of clusters with modes
– Using new dissimilarity measures to deal with
categorical objects
– Using a frequency-based method to update modes of
clusters
– A mixture of categorical and numerical data: k-prototype
method
Data Warehousing/Mining 2
 DBMiner Examples

 Show help on clustering

 Show examples

Data Warehousing/Mining 2
 The K-Medoids Clustering Method

 Find representative objects, called medoids, in clusters

 PAM (Partitioning Around Medoids, 1987)
– starts from an initial set of medoids and iteratively replaces one of
the medoids by one of the non-medoids if it improves the total
distance of the resulting clustering
– PAM works effectively for small data sets, but does not scale well
for large data sets
 CLARA (Kaufmann & Rousseeuw, 1990)
 CLARANS (Ng & Han, 1994): Randomized sampling
 Focusing + spatial data structure (Ester et al., 1995)

Data Warehousing/Mining 3
 k-medoids algorithm

 Use real object to represent the cluster

– Select k representative objects arbitrarily
– repeat
 Assign each remaining object to the cluster of the nearest
medoid
 Randomly select a nonmedoid object
 Compute the total cost, S, of swapping oj with orandom
 If S < 0 then swap oj with orandom
– until there is no change

Data Warehousing/Mining 3
 K-Medoids example

 1, 2, 6, 7, 8, 10, 15, 17, 20 – break into 3 clusters

– Cluster = 6 – 1, 2
– Cluster = 7
– Cluster = 8 – 10, 15, 17, 20
 Random non-medoid – 15 replace 7 (total cost=-13)
– Cluster = 6 – 1 (cost 0), 2 (cost 0), 7(1-0=1)
– Cluster = 8 – 10 (cost 0)
– New Cluster = 15 – 17 (cost 2-9=-7), 20 (cost 5-12=-7)
 Replace medoid 7 with new medoid (15) and reassign
– Cluster = 6 – 1, 2, 7
– Cluster = 8 – 10
– Cluster = 15 – 17, 20
Data Warehousing/Mining 3
 K-Medoids example (continued)
 Random non-medoid – 1 replaces 6 (total cost=2)
– Cluster = 8 – 7 (cost 6-1=5)10 (cost 0)
– Cluster = 15 – 17 (cost 0), 20 (cost 0)
– New Cluster = 1 – 2 (cost 1-4=-3)
 2 replaces 6 (total cost=1)
 Don’t replace medoid 6
– Cluster = 6 – 1, 2, 7
– Cluster = 8 – 10
– Cluster = 15 – 17, 20
 Random non-medoid – 7 replaces 6 (total cost=2)
– Cluster = 8 – 10 (cost 0)
– Cluster = 15 – 17(cost 0), 20(cost 0)
– New Cluster = 7 – 6 (cost 1-0=1), 2 (cost 5-4=1)
Data Warehousing/Mining 3
 K-Medoids example (continued)
 Don’t Replace medoid 6
– Cluster = 6 – 1, 2, 7
– Cluster = 8 – 10
– Cluster = 15 – 17, 20
 Random non-medoid – 10 replaces 8 (total cost=2) don’t
replace
– Cluster = 6 – 1(cost 0), 2(cost 0), 7(cost 0)
– Cluster = 15 – 17 (cost 0), 20(cost 0)
– New Cluster = 10 – 8 (cost 2-0=2)
 Random non-medoid – 17 replaces 15 (total cost=0) don’t
replace
– Cluster = 6 – 1(cost 0), 2(cost 0), 7(cost 0)
– Cluster = 8 – 10 (cost 0)
– New Cluster = 17 – 15 (cost 2-0=2), 20(cost 3-5=-2)
Data Warehousing/Mining 3
 K-Medoids example (continued)
 Random non-medoid – 20 replaces 15 (total cost=3)
don’t replace
– Cluster = 6 – 1(cost 0), 2(cost 0), 7(cost 0)
– Cluster = 8 – 10 (cost 0)
– New Cluster = 20 – 15 (cost 5-0=2), 17(cost 3-2=1)
 Other possible changes all have high costs
– 1 replaces 15, 2 replaces 15, 1 replaces 8, …
 No changes, final clusters
– Cluster = 6 – 1, 2, 7
– Cluster = 8 – 10
– Cluster = 15 – 17, 20
Data Warehousing/Mining 3
 Hierarchical Clustering
 Use distance matrix as clustering criteria. This method
does not require the number of clusters k as an input, but
needs a termination condition
Step 0 Step 1 Step 2 Step 3 Step 4
agglomerative
(AGNES)
a
ab Bottom-up
Places each object in its own
b abcde cluster & then merges
c
cde
d Top-down
de All objects in a single
e cluster & then subdivides
divisive
Step 4 Step 3 Step 2 Step 1 Step 0 (DIANA)
Data Warehousing/Mining 3
 A Dendrogram Shows How the
Clusters are Merged Hierarchically

Decompose data objects into a several levels of nested

partitioning (tree of clusters), called a dendrogram.

A clustering of the data objects is obtained by cutting the

dendrogram at the desired level, then each connected
component forms a cluster.

Data Warehousing/Mining 3
 More on Hierarchical Clustering Methods

 Major weakness of agglomerative clustering methods

– do not scale well: time complexity of at least O(n2), where n is the
number of total objects
– can never undo what was done previously
 Integration of hierarchical with distance-based clustering
– BIRCH (1996): uses CF-tree and incrementally adjusts the quality of
sub-clusters
– CURE (1998): selects well-scattered points from the cluster and
then shrinks them towards the center of the cluster by a specified
fraction
– CHAMELEON (1999): hierarchical clustering using dynamic
modeling

Data Warehousing/Mining 3
 BIRCH (1996)
 Birch: Balanced Iterative Reducing and Clustering using
Hierarchies, by Zhang, Ramakrishnan, Livny (SIGMOD’96)
 Incrementally construct a CF (Clustering Feature) tree, a
hierarchical data structure for multiphase clustering
– Phase 1: scan DB to build an initial in-memory CF tree (a multi-level
compression of the data that tries to preserve the inherent clustering
structure of the data)
– Phase 2: use an arbitrary clustering algorithm to cluster the leaf
nodes of the CF-tree
 Scales linearly: finds a good clustering with a single scan and
improves the quality with a few additional scans
 Weakness: handles only numeric data, and sensitive to the
order of the data record.
Data Warehousing/Mining 3
 CURE (Clustering Using
REpresentatives )

 CURE: proposed by Guha, Rastogi & Shim, 1998

– Stops the creation of a cluster hierarchy if a level
consists of k clusters
– Uses multiple representative points to evaluate the
distance between clusters, adjusts well to arbitrary
shaped clusters and avoids single-link effect
Data Warehousing/Mining 4
 Drawbacks of Distance-Based
Method

 Drawbacks of square-error based clustering method

– Consider only one point as representative of a cluster
– Good only for convex shaped, similar size and density, and if k
can be reasonably estimated

Data Warehousing/Mining 4
 Cure: The Algorithm

– Draw random sample s.

– Partition sample to p partitions with size s/p
– Partially cluster partitions into s/pq clusters
– Eliminate outliers
 By random sampling
 If a cluster grows too slow, eliminate it.
– Cluster partial clusters.
– Label data in disk
Data Warehousing/Mining 4
 CHAMELEON
 CHAMELEON: hierarchical clustering using dynamic
modeling, by G. Karypis, E.H. Han and V. Kumar’99
 Measures the similarity based on a dynamic model
– Two clusters are merged only if the interconnectivity and
closeness (proximity) between two clusters are high relative to the
internal interconnectivity of the clusters and closeness of items
within the clusters
 A two phase algorithm
– 1. Use a graph partitioning algorithm: cluster objects into a
large number of relatively small sub-clusters
– 2. Use an agglomerative hierarchical clustering algorithm: find
the genuine clusters by repeatedly combining these sub-
clusters
Data Warehousing/Mining 4
 Overall Framework of
CHAMELEON
Construct
Sparse Graph Partition the Graph

Data Set

Merge Partition

Final Clusters

Data Warehousing/Mining 4
 Density-Based Clustering
Methods
 Clustering based on density (local cluster criterion),
such as density-connected points
 Major features:
– Discover clusters of arbitrary shape
– Handle noise
– One scan
– Need density parameters as termination condition
 Several interesting studies:
– DBSCAN: Ester, et al. (KDD’96)
– OPTICS: Ankerst, et al (SIGMOD’99).
– DENCLUE: Hinneburg & D. Keim (KDD’98)
– CLIQUE: Agrawal, et al. (SIGMOD’98)

Data Warehousing/Mining 4
 Density-Based Clustering: Background
 Two parameters:
– Epsilon (Eps): Maximum radius of the neighbourhood
– MinPts: Minimum number of points in an Epsilon-neighbourhood
of that point
 NEps(p): {q belongs to D | dist(p,q) <= Eps}
 Directly density-reachable: A point p is directly density-
reachable from a point q wrt. Eps, MinPts if
– 1) p belongs to NEps(q)
– 2) core point condition:
p MinPts = 5
|NEps (q)| >= MinPts
q
Eps = 1 cm

Data Warehousing/Mining 4
 Density-Based Clustering: Background
(II)
 Density-reachable:
p
– A point p is density-reachable from a
point q wrt. Eps, MinPts if there is a chain p1
of points p1, …, pn, p1 = q, pn = p such that q
pi+1 is directly density-reachable from pi

 Density-connected
– A point p is density-connected to a point p q
q wrt. Eps, MinPts if there is a point o
such that both, p and q are density-
o
reachable from o wrt. Eps and MinPts.

Data Warehousing/Mining 4
 DBSCAN: Density Based Spatial
Clustering of Applications with Noise
 Relies on a density-based notion of cluster: A cluster is
defined as a maximal set of density-connected points
 Discovers clusters of arbitrary shape in spatial databases
with noise

Outlier

Border
Eps = 1cm
Core MinPts = 5

Data Warehousing/Mining 4
 DBSCAN: The Algorithm
– Arbitrary select a point p

– Retrieve all points density-reachable from p wrt Eps

and MinPts.
– If p is a core point, a cluster is formed.

– If p is a border point, no points are density-reachable

from p and DBSCAN visits the next point of the
database.
– Continue the process until all of the points have been
processed.
Data Warehousing/Mining 4
 OPTICS: A Cluster-Ordering Method
(1999)
 OPTICS: Ordering Points To Identify the
Clustering Structure
– Ankerst, Breunig, Kriegel, and Sander (SIGMOD’99)
– Produces a special order of the database wrt its
density-based clustering structure
– This cluster-ordering contains info equiv to the
density-based clusterings corresponding to a broad
range of parameter settings
– Good for both automatic and interactive cluster
analysis, including finding intrinsic clustering
structure
– Can be represented graphically or using visualization
techniques
Data Warehousing/Mining 5
 OPTICS: Some Extension from
DBSCAN
 Index-based:
 k = number of dimensions

N = 20
 p = 75% D
 M = N(1-p) = 5
– Complexity: O(kN2)
 Core Distance
p1
 Reachability Distance o
p2
o
Max (core-distance (o), d (o, p))
MinPts = 5
r(p1, o) = 2.8cm. r(p2,o) = 4cm
Data Warehousing/Mining  = 3 cm 5
 Reachability-distance

undefined


 ‘

Cluster-order
Data Warehousing/Mining of the objects 5
 DENCLUE: using density functions
 DENsity-based CLUstEring by Hinneburg & Keim
(KDD’98)
 Major features
– Solid mathematical foundation
– Good for data sets with large amounts of noise
– Allows a compact mathematical description of
arbitrarily shaped clusters in high-dimensional data sets
– Significant faster than existing algorithm (faster than
DBSCAN by a factor of up to 45)
– But needs a large number of parameters
Data Warehousing/Mining 5
 Denclue: Technical Essence
 Uses grid cells but only keeps information about grid cells
that do actually contain data points and manages these cells
in a tree-based access structure.
 Influence function: describes the impact of a data point
within its neighborhood.
 Overall density of the data space can be calculated as the
sum of the influence function of all data points.
 Clusters can be determined mathematically by identifying
density attractors.
 Density attractors are local maximal of the overall density
function.

Data Warehousing/Mining 5
 Gradient: The steepness of a slope

 Example
d ( x , y )2

f Gaussian ( x , y )  e 2 2

d ( x , xi ) 2

( x )  i 1 e
D N
2 2
f Gaussian
d ( x , xi ) 2

( x, xi )  i 1 ( xi  x)  e
N
f D
Gaussian
2 2

Data Warehousing/Mining 5
 Density Attractor

Data Warehousing/Mining 5
 Center-Defined and Arbitrary

Data Warehousing/Mining 5
 Grid-Based Clustering Method
 Using multi-resolution grid data structure
 Several interesting methods
– STING (a STatistical INformation Grid approach)
by Wang, Yang and Muntz (1997)
– WaveCluster by Sheikholeslami, Chatterjee, and
Zhang (VLDB’98)
 A multi-resolution clustering approach using
wavelet method
– CLIQUE: Agrawal, et al. (SIGMOD’98)

Data Warehousing/Mining 5
 STING: A Statistical Information
Grid Approach
 Wang, Yang and Muntz (VLDB’97)
 The spatial area area is divided into rectangular cells
 There are several levels of cells corresponding to different
levels of resolution

Data Warehousing/Mining 5
 STING: A Statistical Information
Grid Approach (2)
– Each cell at a high level is partitioned into a number of smaller cells in
the next lower level
– Statistical info of each cell is calculated and stored beforehand and is
used to answer queries
– Parameters of higher level cells can be easily calculated from parameters
of lower level cell
 count, mean, s, min, max

 type of distribution—normal, uniform, etc.

– Use a top-down approach to answer spatial data queries

– Start from a pre-selected layer—typically with a small number of cells
– For each cell in the current level compute the confidence interval

Data Warehousing/Mining 6
 STING: A Statistical
Information Grid Approach (3)
– Remove the irrelevant cells from further consideration
– When finish examining the current layer, proceed to the next
lower level
– Repeat this process until the bottom layer is reached
– Advantages:
 Query-independent, easy to parallelize, incremental update

 O(K), where K is the number of grid cells at the lowest level

– Disadvantages:
 All the cluster boundaries are either horizontal or vertical,

and no diagonal boundary is detected

Data Warehousing/Mining 6
 WaveCluster (1998)
 Sheikholeslami, Chatterjee, and Zhang (VLDB’98)
 A multi-resolution clustering approach which applies
wavelet transform to the feature space
– A wavelet transform is a signal processing technique that
decomposes a signal into different frequency sub-band.
 Both grid-based and density-based
 Input parameters:
– # of grid cells for each dimension
– the wavelet, and the # of applications of wavelet transform.

Data Warehousing/Mining 6
 WaveCluster (1998)
 How to apply wavelet transform to find clusters
– Summaries the data by imposing a
multidimensional grid structure onto data space
– These multidimensional spatial data objects are
represented in a n-dimensional feature space
– Apply wavelet transform on feature space to find the
dense regions in the feature space
– Apply wavelet transform multiple times which result
in clusters at different scales from fine to coarse

Data Warehousing/Mining 6
 What Is Wavelet (2)?

Data Warehousing/Mining 6
 Quantization

Data Warehousing/Mining 6
 Transformation

Data Warehousing/Mining 6
 WaveCluster (1998)
 Why is wavelet transformation useful for clustering
– Unsupervised clustering
It uses hat-shape filters to emphasize region where points
cluster, but simultaneously to suppress weaker information in
their boundary
– Effective removal of outliers
– Multi-resolution
– Cost efficiency
 Major features:
– Complexity O(N)
– Detect arbitrary shaped clusters at different scales
– Not sensitive to noise, not sensitive to input order
– Only applicable to low dimensional data
Data Warehousing/Mining 6
 CLIQUE (Clustering In QUEst)
 Agrawal, Gehrke, Gunopulos, Raghavan (SIGMOD’98).
 Automatically identifying subspaces of a high dimensional
data space that allow better clustering than original space
 CLIQUE can be considered as both density-based and grid-
based
– It partitions each dimension into the same number of equal length
interval
– It partitions an m-dimensional data space into non-overlapping
rectangular units
– A unit is dense if the fraction of total data points contained in the
unit exceeds the input model parameter
– A cluster is a maximal set of connected dense units within a
subspace

Data Warehousing/Mining 6
 CLIQUE: The Major Steps
 Partition the data space and find the number of points that
lie inside each cell of the partition.
 Identify the subspaces that contain clusters using the
Apriori principle
 Identify clusters:
– Determine dense units in all subspaces of interests
– Determine connected dense units in all subspaces of interests.
 Generate minimal description for the clusters
– Determine maximal regions that cover a cluster of connected
dense units for each cluster
– Determination of minimal cover for each cluster
Data Warehousing/Mining 7
 Strength and Weakness of CLIQUE
 Strength
– It automatically finds subspaces of the highest
dimensionality such that high density clusters exist in
those subspaces
– It is insensitive to the order of records in input and
does not presume some canonical data distribution
– It scales linearly with the size of input and has good
scalability as the number of dimensions in the data
increases
 Weakness
– The accuracy of the clustering result may be degraded
at the expense of simplicity of the method
Data Warehousing/Mining 7
 Model-Based Clustering Methods
 Attempt to optimize the fit between the data and
some mathematical model
 Statistical and AI approach
– Conceptual clustering
 A form of clustering in machine learning
 Produces a classification scheme for a set of unlabeled objects
 Finds characteristic description for each concept (class)
– COBWEB (Fisher’87)
 A popular a simple method of incremental conceptual learning
 Creates a hierarchical clustering in the form of a classification
tree
 Each node refers to a concept and contains a probabilistic
description of that concept

Data Warehousing/Mining 7
 More on Statistical-Based
Clustering
 Limitations of COBWEB
– The assumption that the attributes are independent of each
other is often too strong because correlation may exist
– Not suitable for clustering large database data – skewed tree
and expensive probability distributions
 CLASSIT
– an extension of COBWEB for incremental clustering of
continuous data
– suffers similar problems as COBWEB
 AutoClass (Cheeseman and Stutz, 1996)
– Uses Bayesian statistical analysis to estimate the number of
clusters
– Popular in industry

Data Warehousing/Mining 7
 Other Model-Based
Clustering Methods
 Neural network approaches
– Represent each cluster as an exemplar, acting as a
“prototype” of the cluster
– New objects are distributed to the cluster whose
exemplar is the most similar according to some
dostance measure
 Competitive learning
– Involves a hierarchical architecture of several units
(neurons)
– Neurons compete in a “winner-takes-all” fashion
for the object currently being presented
Data Warehousing/Mining 7
 Self-organizing feature maps
(SOMs)
 Clustering is also performed by having several
units competing for the current object
 The unit whose weight vector is closest to the
current object wins
 The winner and its neighbors learn by having
their weights adjusted
 SOMs are believed to resemble processing that
can occur in the brain
 Useful for visualizing high-dimensional data in
2- or 3-D space

Data Warehousing/Mining 7
 What Is Outlier Discovery?
 What are outliers?
– The set of objects are considerably dissimilar from
the remainder of the data
– Example: Sports: Michael Jordon, Wayne Gretzky, ...
 Problem
– Find top n outlier points
 Applications:
– Credit card fraud detection
– Telecom fraud detection
– Customer segmentation
– Medical analysis
Data Warehousing/Mining 7
 Outlier Discovery:
Statistical
Approaches

 Assume a model underlying distribution that

generates data set (e.g. normal distribution)
 Use discordancy tests depending on
– data distribution
– distribution parameter (e.g., mean, variance)
– number of expected outliers
 Drawbacks
– most tests are for single attribute
– In many cases, data distribution may not be known
Data Warehousing/Mining 7
 Outlier Discovery: Distance-
Based Approach
 Introduced to counter the main limitations imposed by
statistical methods
– We need multi-dimensional analysis without knowing data
distribution.
 Distance-based outlier: A Distane Based(p, D)-outlier is
an object O in a dataset T such that at least a fraction p of
the objects in T lies at a distance greater than D from O
 Algorithms for mining distance-based outliers
– Index-based algorithm
– Nested-loop algorithm
– Cell-based algorithm

Data Warehousing/Mining 7
 Outlier Discovery: Deviation-
Based Approach
 Identifies outliers by examining the main characteristics
of objects in a group
 Objects that “deviate” from this description are
considered outliers
 sequential exception technique
– simulates the way in which humans can distinguish unusual
objects from among a series of supposedly like objects
 OLAP data cube technique
– uses data cubes to identify regions of anomalies in large
multidimensional data

Data Warehousing/Mining 7
 Problems and Challenges
 Considerable progress has been made in scalable
clustering methods
– Partitioning: k-means, k-medoids, CLARANS
– Hierarchical: BIRCH, CURE
– Density-based: DBSCAN, CLIQUE, OPTICS
– Grid-based: STING, WaveCluster
– Model-based: Autoclass, Denclue, Cobweb
 Current clustering techniques do not address all the
requirements adequately
 Constraint-based clustering analysis: Constraints exist in
data space (bridges and highways) or in user queries
Data Warehousing/Mining 8
 Constraint-Based Clustering Analysis

 Clustering analysis: less parameters but more user-desired

constraints, e.g., an ATM allocation problem

Data Warehousing/Mining 8
 Summary
 Cluster analysis groups objects based on their similarity
and has wide applications
 Measure of similarity can be computed for various types
of data
 Clustering algorithms can be categorized into partitioning
methods, hierarchical methods, density-based methods,
grid-based methods, and model-based methods
 Outlier detection and analysis are very useful for fraud
detection, etc. and can be performed by statistical,
distance-based or deviation-based approaches
 There are still lots of research issues on cluster analysis,
such as constraint-based clustering
Data Warehousing/Mining 8