Fuzzy Decision Forest

Author

Janikow, Cezary Z.

Author_Institution

Dept. of Math. & Comput. Sci., Missouri Univ., St. Louis, MO, USA

fYear

2003

fDate

24-26 July 2003

Firstpage

480

Lastpage

483

Abstract

In the past, we have developed and presented a Fuzzy Decision Tree, more recently followed by an extension called a Fuzzy Decision Forest. The idea behind the forest is not only to represent multiple trees, but also to represent test alternatives at all levels of every tree. The resulting tree is in fact a 3-dimensional tree. A two-dimensional slice is equivalent to a single decision tree. The forest allows multiple choices of tests in some or all nodes of the decision tree. These alternative tests can be used to enhance the classification accuracy of the tree. However, the major advantage of having multiple test choices is to have alternative test decisions when features in test data are unreliable or just missing. In the paper, we overview the ideas behind Fuzzy Decision Forest, and we illustrate its enhanced capabilities with a number of experiments with missing features.

Keywords

decision theory; decision trees; fuzzy set theory; pattern classification; 3-dimensional tree; classification accuracy; fuzzy decision forest; fuzzy decision tree; multiple test choice; multiple trees; single decision tree; test data features; test decisions; two-dimensional slice; Binary trees; Classification tree analysis; Computer science; Data mining; Decision trees; Fuzzy sets; Robust stability; Testing; Training data; Working environment noise;

fLanguage

English

Publisher

ieee

Conference_Titel

Fuzzy Information Processing Society, 2003. NAFIPS 2003. 22nd International Conference of the North American

Print_ISBN

0-7803-7918-7

Type

conf

DOI

10.1109/NAFIPS.2003.1226832

Filename

1226832