Author(s)

W. F. Castilho, H. A. do Prado & M. Ladeira

Abstract

Introducing prior knowledge into the clustering process W. F. ~astilho'>~, H. A. do pradola3 & M. ~adeira~ I Catholic University of Brasilia, Brazil 2 Federal Savings Bank, Brazil 3 Brazilian Enterprise for Agricultural Research, Brazil 3 University of Brasilia, Brazil Abstract A cluster is a subset of elements from a population, defined as a high-density region separated from other clusters by low-density regions. By applying a clustering procedure, an analyst looks for a description about how data is structured in its dimension space. Then, the structure is explored in order to figure out interesting groups that can, eventually, be associated to classes. Since the late 90s, a significant effort has been undertaken to address the clustering approach for data mining. However, the contributions have focused mainly on algorithm optimization. An important aspect of the clustering process, namely, the use of prior knowledge to constrain the clusters definition, has rarely been contemplated. Accor

Keywords