This preview shows page 1. Sign up to view the full content.
Unformatted text preview: ls, how many beds should be provided in different
sections of the hospital, what types of medicines and in how much quantity should
be stocked, and so on.
The objective of clustering algorithms for data mining is to automatically partition
the data space of a database into a set of clusters, to which each of the objects in
the database are assigned, either deterministically or probabilistically. A clustering
algorithm is designed to identify all sets of similar objects in a database, in some
optimal fashion. The objects are so clustered that the intra-cluster (objects
belonging to the same cluster) similarities are maximized, and the inter-cluster
(objects belonging to different clusters) similarities are minimized. This is done
based on some criteria defined on the attributes of the objects. Once the clusters
are decided, the common features for objects in a cluster are summarized to form
the cluster description.
It is important to understand the difference between classification and clustering
algorithms. In the case of classification, the objects of a database are classified
into predefined classes whose number is known and fixed before the classification
algorithm is applied to the database. However, in case of clustering, the objects of
a database are clustered into clusters whose number is not known before the
clustering algorithm is applied to the database. Thus, if the same classification
algorithm is applied to two different databases, both will result in the same
predefined number of classes (say n, some of which may be null if no object mans
to them). On the other hand, if the same clustering algorithm is applied to two
different databases, both may result in different, previously unknown number of clusters (say m and ri). Also note that the common features of objects in a class are
pre-defined in the classification algorithm, but the common features of objects in a
cluster are summarized after all the objects have been put in the cluster.
A simple application of clustering algorithm may be to cluster the customers of a
bank from the ban...
View Full Document
This document was uploaded on 04/07/2014.
- Spring '14