|
Combine Clustering and Association Rules
Traditionally,
clustering and association rule mining are two independent data mining
tasks. We propose to combine them, i.e., first clustering the
database, and then mining association rules in the clusters.
Benefit:
- Clustering
will have a better cluster profile by generating association rules in
each cluster.
- The
association rules generated from the clusters will be more significant
and more meaningful than those found from the whole database. Such
significant and meaningful association rules, however, may not be
found in the whole database unless the minimum support is set
extremely low.
Online
and Adaptive:
We also
plan to bring the comibination of clustering and association rule mining
both online and adaptive, that is, the user is shown continuously the current
association rules in the current clusters, based on these the user can
adjust the parameters(k, k' for clustering, minimum support and confidence
for association rules).
|
