8 Partitioning Large Database Using Irrelevant Attributes for Knowledge Discovery and Learning
Download citation file:
- Ris (Zotero)
- Reference Manager
Learning systems are very effective tools for discovering knowledge from databases. Most symbolic learning systems face difficulties when discovering knowledge from large databases. Adapting the learning systems to cope with large databases is the traditional solution to overcome these limitations. This paper evaluates a new methodology for partitioning the representation space. The method selects an irrelevant attribute, determined by a utility function, to partition the representation space. Since irrelevant attributes are not needed to describe concepts learned from the original data, the knowledge discovered from all subspaces should be independent of such an attribute. If the representation space is partitioned by a relevant attribute, the knowledge discovered from all subspaces are combined simply using information about that attribute. The method is analyzed using C5 learning system, or See5.