Automatic Feature Scaling and Selection for Support Vector Machine Classi cation with Functional Data
Author
dc.contributor.author
Jiménez Cordero, Asunción
Author
dc.contributor.author
Maldonado Alarcón, Sebastián
Admission date
dc.date.accessioned
2020-10-19T16:43:46Z
Available date
dc.date.available
2020-10-19T16:43:46Z
Publication date
dc.date.issued
2020
Cita de ítem
dc.identifier.citation
Applied Intelligence (2020)
es_ES
Identifier
dc.identifier.other
10.1007/s10489-020-01765-6
Identifier
dc.identifier.uri
https://repositorio.uchile.cl/handle/2250/177225
Abstract
dc.description.abstract
Functional Data Analysis (FDA) has become a very important eld
in recent years due to its wide range of applications. However, there are several
real-life applications in which hybrid functional data appear, i.e., data with
functional and static covariates. The classi cation of such hybrid functional
data is a challenging problem that can be handled with the Support Vector
Machine (SVM). Moreover, the selection of the most informative features
may yield to drastic improvements in the classi cation rates. In this paper,
an embedded feature selection approach for SVM classi cation is proposed, in
which the isotropic Gaussian kernel is modi ed by associating a bandwidth
to each feature. The bandwidths are jointly optimized with the SVM parameters,
yielding an alternating optimization approach. The e ectiveness of our
methodology was tested on benchmark data sets. Indeed, the proposed method
achieved the best average performance when compared to 17 other feature selection
and SVM classi cation approaches. A comprehensive sensitivity analysis
of the parameters related to our proposal was also included, con rming
its robustness.
es_ES
Patrocinador
dc.description.sponsorship
Spanish Government
MTM2015-65915-R
Junta de Andalucia
P11-FQM-7603
P18-FR-2369
FQM329
German Research Foundation (DFG)
VI PPITUS (Universidad de Sevilla)
EU ERDF funds
FBBVA-COSECLA
ANID, FONDECYT project
1200221
Complex Engineering Systems Institute (ANID, PIA)
FB0816