|
Title:
|
A METHOD FOR COMBINING INSTANCE SELECTION ALGORITHMS |
|
Author(s):
|
Yoel Caises , Antonio González , Enrique Leyva , Raúl Pérez |
|
ISBN:
|
978-972-8924-87-4 |
|
Editors:
|
António Palma dos Reis |
|
Year:
|
2009 |
|
Edition:
|
Single |
|
Keywords:
|
Instance selection, data mining, data reduction, machine learning. |
|
Type:
|
Full Paper |
|
First Page:
|
77 |
|
Last Page:
|
84 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
The goal of instance selection is to improve the efficiency of inductive learning by reducing the data sets before their
processing by learners. This database filtering can also eliminate instances that are harmful to the learning process, such as
those considered to be noise. Many authors have worked in this area and many different algorithms have been proposed in
the literature. However, several studies have shown that no single algorithm is better than all the others over a wide range of
domains with different characteristics.
This paper presents a set of measures to characterize the domains, as well as a new algorithm that uses these measures to
characterize domains and, depending on the characteristics detected, applies the method or combination of methods expected
to produce the best results. An experimental study in which twenty databases are processed by this and five well-known
state-of-the-art methods is also presented. The results of this study were supplied to a fuzzy rule learning algorithm and a
comparison was carried out involving the rate of reduction and classification successes. |
|
|
|
|
|
|
