English español

Repositorio de la Universidad de Oviedo. > Producción Bibliográfica de UniOvi: RECOPILA > Artículos >

Please use this identifier to cite or link to this item: http://hdl.handle.net/10651/25317

Title: Bootstrap analysis of multiple repetitions of experiments using an interval-valued multiple comparison procedure
Author(s): Otero Rodríguez, José
Sánchez Ramos, Luciano
Couso Blanco, Inés
Palacios Jiménez, Ana María
Keywords: Cross validation
Statistical comparisons of algorithms
Tests for interval-valued data
Issue date: 2014
Publisher: Elsevier
Publisher version: http://dx.doi.org/10.1016/j.jcss.2013.03.009
Citation: Journal of Computer and System Sciences, 80(1), p. 88-100 (2014); doi:10.1016/j.jcss.2013.03.009
Format extent: p. 88-100
Abstract: A new bootstrap test is introduced that allows for assessing the signi cance of the di erences between stochastic algorithms in a cross-validation with repeated folds experimental setup. Intervals are used for modeling the variability of the data that can be attributed to the repetition of learning and testing stages over the same folds in cross validation. Numerical experiments are provided that support the following three claims (1) Bootstrap tests can be more powerful than ANOVA or Friedman test for comparing multiple classi ers (2) In the presence of outliers, interval-valued bootstrap tests achieve a better discrimination between stochastic algorithms than nonparametric tests and (3) Choosing ANOVA, Friedman or Bootstrap can produce di erent conclusions in experiments involving actual data from machine learning tasks
URI: http://hdl.handle.net/10651/25317
ISSN: 0022-0000
Local identifier: 20141178
Sponsored: This work has been funded by Spanish Ministry of Economy and Competitiveness, grant TIN2011-24302
Project id.: MEC/TIN2011-24302
Appears in Collections:Artículos
Investigaciones y Documentos OpenAIRE

Files in This Item:

File Description SizeFormat
JCSS.pdfPostprint375,27 kBAdobe PDFView/Open

Exportar a Mendeley

This item is licensed under a Creative Commons License
Creative Commons

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.


Base de Datos de Autoridades Biblioteca Universitaria Consultas / Sugerencias