Ponencia
Deleting or Keeping Outliers for Classifier Training?
Autor/es | Tallón Ballesteros, Antonio Javier
Riquelme Santos, José Cristóbal |
Departamento | Universidad de Sevilla. Departamento de Lenguajes y Sistemas Informáticos |
Fecha de publicación | 2014 |
Fecha de depósito | 2016-06-23 |
Publicado en |
|
ISBN/ISSN | 978-1-4799-5936-5 |
Resumen | This paper introduces two statistical outlier
detection approaches by classes. Experiments on binary and
multi-class classification problems reveal that the partial
removal of outliers improves significantly one or ... This paper introduces two statistical outlier detection approaches by classes. Experiments on binary and multi-class classification problems reveal that the partial removal of outliers improves significantly one or two performance measures for C4.S and I-nearest neighbour classifiers. Also, a taxonomy of problems according to the amount of outliers is proposed. |
Identificador del proyecto | TIN2007- 68084-C02-02
TIN2011-28956-C02-02 Pll-TIC-7528 |
Cita | Tallón Ballesteros, A.J. y Riquelme Santos, J.C. (2014). Deleting or Keeping Outliers for Classifier Training?. En Sixth World Congress on Nature and Biologically Inspired Computing (NaBIC), 2014 (281-286), Porto (Portugal): IEEE. |
Ficheros | Tamaño | Formato | Ver | Descripción |
---|---|---|---|---|
Deletin or keeping.pdf | 564.1Kb | [PDF] | Ver/ | |