Title

Emerging Patterns and Classification

Document Type

Conference Proceeding

Publication Date

11-2000

Find in a Library

Catalog Record

Abstract

In this work, we review an important kind of knowledge pattern, emerging patterns (EPs). Emerging patterns are associated with two data sets, and can be used to describe significant changes between the two data sets. To discover all EPs embedded in high-dimension and large-volume databases is a challenging problem due to the number of candidates. We describe a special type of EP, called jumping emerging patterns (JEPs) and review some properties of JEP spaces (the spaces of jumping emerging patterns). We describe efficient border-based algorithms to derive the boundary elements of JEP spaces. Moreover, we describe a new classifier, called DeEPs, which makes use of the discriminating power of emerging patterns. The experimental results show that the accuracy of DeEPs is much better than that of k-nearest neighbor and that of C5.0.

Comments

Presented at the 6th Asian Computing Science Conference, Penang, Malaysia, November 25-27, 2000.

DOI

10.1007/3-540-44464-5_3

Catalog Record

Share

COinS