Knowledge of clustering changes in real-life datasets
is important in many contexts, such as customer attrition analysis and fraud detection. Organizations
can use such knowledge of change to adapt business
strategies in response to changing circumstances. Analysts should be able to relate new knowledge acquired from a newer dataset to that acquired from
an earlier dataset to understand what has changed.
There are two kind of clustering changes, which are:
changes in clustering structure and changes in cluster memberships. The key contribution of this paper
is a novel method to automatically detect structural
changes in two snapshot datasets using ReDSOM.
The method identifies emerging clusters, disappearing
clusters, splitting clusters, merging clusters, enlarging clusters, and shrinking clusters. Evaluation using synthetic datasets demonstrates that this method
can identify automatically structural cluster changes.
Moreover, the changes identified in our evaluation using real-life datasets from the World Bank can be related to actual changes.
|Cite as: Denny, Wicaksono, P. and Manurung, R. (2014). Automatic Detection of Cluster Structure Changes using Relative Density Self-Organizing Maps. In Proc. Twelfth Australasian Data Mining Conference (AusDM14) Brisbane, Australia. CRPIT, 158. Li, X., Liu, L., Ong, K.L. and Zhao, Y. Eds., ACS. 9-17 |
(local if available)