Chapter 3: Cluster Analysis
3.1 Basic Concepts of Clustering
3.1.1 Cluster Analysis
3.1.2 Clustering Categories
3.2 Partitioning Methods
3.2.1 The principle
3.2.2 K-Means Method
3.2.3 K-Medoids Method
3.2.4 CLARA
3.2.5 CLARANS
3.3 Hierarchical Methods
3.4 Density-based Methods
3.5 Clustering High-Dimensional Data
3.6 Outlier Analysis
3.1.1 Cluster Analysis
`
Unsupervised learning (i.e., Class label is unknown)
Group data to form new categories (i.e., clusters), e.g., cluster
houses to find distribution patterns
Principle: Maximizing intra-class similarity & minimizing interclass
similarity
Typical Applications
WWW, Social networks, Marketing, Biology, Library, etc.
3.1.2 Clustering Categories
`
Partitioning Methods
Hierarchical Methods
Hypothesize a model for each cluster and find the best fit of the
data to the given model
Clustering high-dimensional data
Quantize the object space into a finite number of cells
Model-based methods
Grow a given cluster depending on its density (# data objects)
Grid-based Methods
Creates a hierarchical decomposition of the data
Density-based Methods
Construct k partitions of the data
Subspace clustering
Constraint-based methods
Used for user-specific applications
Chapter 3: Cluster Analysis
`
3.1 Basic Concepts of Clustering
3.1.1 Cluster Analysis
3.1.2 Clustering Categories
3.2 Partitioning Methods
3.2.1 The principle
3.2.2 K-Means Method
3.2.3 K-Medoids Method
3.2.4 CLARA
3.2.5 CLARANS
3.3 Hierarchical Methods
3.4 Density-based Methods
3.5 Clustering High-Dimensional Data
3.6 Outlier Analysis
3.2.1 Partitioning Methods: The Principle
`
Given
A data set of n objects
K the number of clusters to form
Organize the objects into k partitions (k<=n) where each partition
represents a cluster
The clusters are formed to optimize an objective partitioning
criterion
Objects within a cluster are similar
Objects of different clusters are dissimilar
3.2.2 K-Means Method
Goal:
create 3 clusters
(partitions)
Choose 3 objects
(cluster centroids)
Assign each object
to the closest centroid
to form Clusters
+
Update cluster
centroids
+
+
K-Means Method
Recompute
Clusters
+
+
+
If Stable centroids,
then stop
+
+
+
K-Means Algorithm
`
Input
K: the number of clusters
D: a data set containing n objects
Output: A set of k clusters
Method:
(1) Arbitrary choose k objects from D as in initial cluster centers
(2) Repeat
(3) Reassign each object to the most similar cluster based on the
mean value of the objects in the cluster
(4) Update the cluster means
(5) Until no change
K-Means Properties
`
The algorithm attempts to determine k partitions that minimize the
square-error function
i1
pC
( p m i)
E: the sum of the squared error for all objects in the data set
P: the data point in the space representing an object
mi: is the mean of cluster Ci
It works well when the clusters are compact clouds that are
rather well separated from one another
K-Means Properties
Advantages
`
K-means is relatively scalable and efficient in processing large
data sets
The computational complexity of the algorithm is O(nkt)
n: the total number of objects
k: the number of clusters
t: the number of iterations
Normally: k<<n and t<<n
Disadvantage
`
Can be applied only when the mean of a cluster is defined
Users need to specify k
K-means is not suitable for discovering clusters with nonconvex
shapes or clusters of very different size
It is sensitive to noise and outlier data points (can influence the
mean value)
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 (Huang98)
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
November 2, 2010
Data Mining: Concepts and Techniques
11
3.2.3 K-Medoids Method
`
Minimize the sensitivity of k-means to outliers
Pick actual objects to represent clusters instead of mean values
Each remaining object is clustered with the representative object
(Medoid) to which is the most similar
The algorithm minimizes the sum of the dissimilarities between
each object and its corresponding reference point
i1
pC
| p o
E: the sum of absolute error for all objects in the data set
P: the data point in the space representing an object
Oi: is the representative object of cluster Ci
K-Medoids Method: The Idea
`
Initial representatives are chosen randomly
The iterative process of replacing representative objects by no
representative objects continues as long as the quality of the
clustering is improved
For each representative Object O
For each non-representative object R, swap O and R
Choose the configuration with the lowest cost
Cost function is the difference in absolute error-value if a current
representative object is replaced by a non-representative object
K-Medoids Method: Example
Data Objects
9
8
A1
A2
7
6
O1
O2
O3
O4
O5
O6
O7
O8
O9
O10
4
10
5
4
7
5
Goal: create two clusters
Choose randmly two medoids
O2 = (3,4)
O8 = (7,4)
K-Medoids Method: Example
Data Objects
9
cluster1
A1
O1
A2
7
6
O2
O3
O4
O5
O6
O7
O8
O9
O10
cluster2
3
4
10
5
4
7
5
Assign each object to the closest representative object
Using L1 Metric (Manhattan), we form the following
clusters
Cluster1 = {O1, O2, O3, O4}
Cluster2 = {O5, O6, O7, O8, O9, O10}
K-Medoids Method: Example
9
Data Objects
cluster1
A1
O1
A2
O2
O3
O4
O5
O6
O7
O8
O9
O10
cluster2
3
4
10
5
4
7
5
Compute the absolute error criterion [for the set of
Medoids (O2,O8)]
| p o |
i1 pCi
| o1 o2 | | o3 o2 | | o4 o2 |
| o5 o8 | | o6 o8 | | o7 o8 | | o9 o8 | | o10 o8 |
K-Medoids Method: Example
9
Data Objects
cluster1
A1
O1
A2
O2
O3
O4
O5
O6
O7
O8
O9
O10
cluster2
3
4
10
5
4
7
5
The absolute error criterion [for the set of Medoids
(O2,O8)]
(3 4 4) (3 1 1 2 2)
20
K-Medoids Method: Example
9
Data Objects
cluster1
A1
O1
A2
O2
O3
O4
O5
O6
O7
O8
O9
O10
cluster2
3
4
10
7
5
Choose a random object O7
Swap O8 and O7
Compute the absolute error criterion [for the set of Medoids
(O2,O7)]
(3 4 4) (2 2 1 3 3)
22
K-Medoids Method: Example
Data Objects
9
cluster1
A1
O1
A2
7
6
O2
O3
O4
O5
O6
O7
O8
O9
O10
cluster2
3
4
10
5
4
7
5
Compute the cost function
Absolute error [for O2,O7] Absolute error [O2,O8]
22 20
S> 0 it is a bad idea to replace O8 by O7
K-Medoids Method
9
Data Objects
cluster1
A1
O1
A2
O2
O3
O4
O5
O6
O7
O8
O9
O10
cluster2
3
4
10
5
4
7
5
In this example, changing the medoid of
cluster 2 did not change the assignments of
objects to clusters.
What are the possible cases when we
replace a medoid by another object?
K-Medoids Method
Cluster 1
Cluster 2
First case
B
B
The assignment of P to A does not change
Representative object
Random Object
Currently P assigned to A
Cluster 1
A
Cluster 2
Second case
p
B
Representative object
Random Object
Currently P assigned to B
P is reassigned to A
K-Medoids Method
Cluster 1
A
Cluster 2
Third case
p
B
B
P is reassigned to the new B
Representative object
Random Object
Currently P assigned to B
Cluster 1
Cluster 2
Fourth case
A
B
p
Representative object
Random Object
Currently P assigned to A
P is reassigned to B
K-Medoids Algorithm(PAM)
PAM : Partitioning Around Medoids
`
`
`
Input
K: the number of clusters
D: a data set containing n objects
Output: A set of k clusters
Method:
(1) Arbitrary choose k objects from D as representative objects (seeds)
(2) Repeat
(3) Assign each remaining object to the cluster with the nearest
representative object
(4) For each representative object Oj
(5) Randomly select a non representative object Orandom
(6) Compute the total cost S of swapping representative object Oj with
Orandom
(7) if S<0 then replace Oj with Orandom
(8) Until no change
K-Medoids Properties(k-medoids vs.K-means)
`
The complexity of each iteration is O(k(n-k)2)
For large values of n and k, such computation becomes very costly
Advantages
K-Medoids method is more robust than k-Means in the presence of
noise and outliers
Disadvantages
K-Medoids is more costly that the k-Means method
Like k-means, k-medoids requires the user to specify k
It does not scale well for large data sets
3.2.4 CLARA
`
CLARA (Clustering Large Applications) uses a sampling-based
method to deal with large data sets
A random sample
should closely represent the
original data
`
The chosen medoids
will likely be similar
to what would have been
chosen from the whole
data set
PAM
sample
CLARA
`
Draw multiple samples of the data set
Apply PAM to each sample
Choose the best clustering
Clusters
Clusters
Clusters
PAM
PAM
PAM
Return the best clustering
sample1
sample2
samplem
CLARA Properties
`
Complexity of each Iteration is: O(ks2 + k(n-k))
s: the size of the sample
k: number of clusters
n: number of objects
PAM finds the best k medoids among a given data, and CLARA
finds the best k medoids among the selected samples
Problems
The best k medoids may not be selected during the sampling
process, in this case, CLARA will never find the best clustering
If the sampling is biased we cannot have a good clustering
Trade off-of efficiency
3.2.5 CLARANS
`
CLARANS (Clustering Large Applications based upon
RANdomized Search ) was proposed to improve the quality and
the scalability of CLARA
It combines sampling techniques with PAM
It does not confine itself to any sample at a given time
It draws a sample with some randomness in each step of the
search
CLARANS: The idea
Clustering view
Current
medoids
Cost=10
Cost=5
Cost=2
Cost=1
Cost=3
Keep the current medoids
medoids
Cost=20
Cost=5
CLARANS: The idea
CLARA
` Draws a sample of nodes at the beginning of the search
` Neighbors are from the chosen sample
` Restricts the search to a specific area of the original data
Sample
medoids
First step of the search
Neighbors are from the chosen sample
second step of the search
Neighbors are from the chosen sample
Current
medoids
CLARANS: The idea
CLARANS
` Does not confine the search to a localized area
` Stops the search when a local minimum is found
` Finds several local optimums and output the clustering
with the best local optimum
First step of the search
Draw a random sample of neighbors
Original data
medoids
Current
medoids
second step of the search
Draw a random sample of neighbors
The number of neighbors sampled from the original data is specified by the user
CLARANS Properties
`
Advantages
Experiments show that CLARANS is more effective than both PAM
and CLARA
Handles outliers
Disadvantages
The computational complexity of CLARANS is O(n2), where n is the
number of objects
The clustering quality depends on the sampling method
Summary of Section 3.2
`
Partitioning methods find sphere-shaped clusters
K- mean is efficient for large data sets but sensitive to outliers
PAM uses centers of the clusters instead of means
CLARA and CLARANS are used for clustering large databases
