Mariana Tasca Fontenelle Lôbo,  Bianca Zadrozny,  Alexandre Plastino. 

An important step in the knowledge discovery in databases (KDD) process is attribute selection, i.e., choosing a subset of attributes that can ade­quately represent the important information that exists in the data. Traditional attribute selection methods do not deal with multi­valued attributes, which are attributes that may assume more than one value for the same instance. This paper proposes a relevance measure for multi­valued attributes, aimed at as­ sessing their importance for classification, which can be used as a criterion for attribute selection. The proposed measure takes into consideration the ability that each attribute has in determining the instance's class. Experimental results with several datasets show that the proposed measure is a good indicator of an attribute's importance for classification.

http://www.lbd.dcc.ufmg.br:8080/colecoes/waamd/2009/007.pdf

Processando consulta.