Gene clustering: 2001:
2
Select
Good Clusters
§
Cluster
=
instances with common features
•
but many clusters irrelevant to experiment
•
many clusters not meaningful
§
Select "pure" clusters
•
At least
x
% of instances have same category
•
x
is user specified (accuracy of rules)
§
Select “significant” clusters
•
At least
m
instances.
•
m
determined by significance test