In this study, a method is presented for selection and
ranking of significant attributes for individual clusters which lead to
formulation of concise and user understandable patterns. Cluster is set
of similar data objects and similarity is measured on attribute values.
Attributes which have same value for majority of objects in a cluster
are considered significant and rest non significant for that cluster.
Reduct from rough set theory is defined as the set of attributes which
distinguishes the objects in a homogenous cluster, therefore these can
be clear cut removed from the same. Non reduct attributes are ranked for
their contribution in the cluster. Pattern is then formed by conjunction
of most contributing attributes of that cluster.