CLUSTER ANALYSIS

|

 
  

    
 O AT IS CLUSTER ANALYSIS

˜ Technique for Data Mining

˜ Classifying items into groupings: clusters
˜ Exploratory data analysis tool
˜ Cluster: a collection of data objects
è Similar to one another within the same cluster
è Dissimilar to the objects in other clusters
- Dinding similarities between data according to the
characteristics found in the data and grouping similar data
objects into clusters
 APPLICATIONS

˜ Draud detection; credit cards & e-commerce

˜ Market segmentation of customers
˜ Pattern Recognition
˜ Spatial Data Analysis
è Create thematic maps in GIS by clustering feature spaces
è Detect spatial clusters or for other spatial mining tasks
 Ohy Clustering in Data Mining?

˜ Scalability
˜ Ability to deal with different types of attributes
˜ Ability to handle dynamic data
˜ 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
˜ igh dimensionality
˜ Incorporation of user-specified constraints
˜ Interpretability and usability
 Ohat Is Good Clustering?

˜ A good clustering method will produce high quality

clusters with
è high intra-class similarity
è low inter-class similarity
Similarity matrix: Similarity is expressed in terms of a distance
function, typically matrix: (`

It measures similarity between every pair of item to be clustered.






 ierarchical Clustering

?
 ?
 ?
 ?
 ?


þ
þ

 



 











?
 ?
 ?
 ?
 ?
 


 Partitioning Clustering

˜ Construct a partition of a database „ of objects

into a set of
clusters, s.t., sum of squared distance
is minimized.
˜ Given a ·, find a partition of ·
that optimizes
the chosen partitioning criterion
˜ The object is to partition a set of Vobjects into a
number ·
( 
  ( clusters by maximizing the
distance between cluster centers while minimizing
the within-cluster variation.