Ricardo Cerri,  André P. L. F. de Carvalho,  Eduardo de P. Costa,  Alex Freitas. 

Most of data classification problems assume that the classes are organized in a flat structure. However, in some domains, like bioinformatics and text mining, classes can be organized in a hierarchical fashion, where classes may have subclasses and superclasses. One such problem is protein function prediction. Machine Learning algorithms have been widely used to induct classification models to predict the functions of new proteins organized hierarchically. This paper investigates the application of the top-down and big-bang classification approaches under different metrics, using decision trees, in hierarchical classification problems.

http://www.lbd.dcc.ufmg.br:8080/colecoes/waamd/2008/006.pdf

Processando consulta.