An implementation of feature selection and ranking via simultaneous perturbation
stochastic approximation (SPSA-FSR) based on works by V. Aksakalli and M. Malekipirbazari
(2015) <arXiv:1508.07630> and Zeren D. Yenice and et al. (2018) <arXiv:1804.05589>.
The SPSA-FSR algorithm searches for a locally optimal set of features that yield the best
predictive performance using a specified error measure such as mean squared error (for
regression problems) and accuracy rate (for classification problems). This package requires
an object of class 'task' and an object of class 'Learner' from the 'mlr' package.
| Version: |
1.0.0 |
| Depends: |
mlr (≥ 2.11), parallelMap (≥ 1.3), parallel (≥ 3.4.2), tictoc (≥ 1.0) |
| Imports: |
ggplot2 (≥ 2.2.1), class (≥ 7.3), mlbench (≥ 2.1) |
| Suggests: |
caret (≥ 6.0), MASS (≥ 7.3), knitr, rmarkdown |
| Published: |
2018-05-11 |
| Author: |
Vural Aksakalli [aut, cre],
Babak Abbasi [aut, ctb],
Yong Kai Wong [aut, ctb],
Zeren D. Yenice [ctb] |
| Maintainer: |
Vural Aksakalli <vaksakalli at gmail.com> |
| BugReports: |
https://github.com/yongkai17/spFSR/issues |
| License: |
GPL-3 |
| URL: |
https://www.featureranking.com/, https://arxiv.org/abs/1804.05589 |
| NeedsCompilation: |
no |
| CRAN checks: |
spFSR results |