OSTSC: Over Sampling for Time Series Classification

Oversampling of imbalanced univariate time series classification data using integrated ESPO and ADASYN methods. Enhanced Structure Preserving Oversampling (ESPO) is used to generate a large percentage of the synthetic minority samples from univariate labeled time series under the modeling assumption that the predictors are Gaussian. ESPO estimates the covariance structure of the minority-class samples and applies a spectral filer to reduce noise. Adaptive Synthetic (ADASYN) sampling approach is a nearest neighbor interpolation approach which is subsequently applied to the ESPO samples. This code is ported from a 'MATLAB' implementation by Cao et al. <doi:10.1109/TKDE.2013.37> and adapted for use with Recurrent Neural Networks implemented in 'TensorFlow'.

Version: 0.0.1
Depends: R (≥ 3.2.3)
Imports: fields, MASS, stats, utils, parallel, doParallel, doSNOW, foreach
Suggests: knitr, rmarkdown, keras, dummies, rlist, pROC, devtools, knitcitations, testthat, xts
Published: 2017-12-04
Author: Matthew Dixon [ctb], Diego Klabjan [ctb], Lan Wei [aut, trl, cre]
Maintainer: Lan Wei <lweicdsor at gmail.com>
License: GPL-3
URL: https://github.com/lweicdsor/OSTSC
NeedsCompilation: no
CRAN checks: OSTSC results


Reference manual: OSTSC.pdf
Vignettes: Over_Sampling_for_Time_Series_Classification
Package source: OSTSC_0.0.1.tar.gz
Windows binaries: r-devel: OSTSC_0.0.1.zip, r-release: OSTSC_0.0.1.zip, r-oldrel: OSTSC_0.0.1.zip
OS X binaries: r-release: OSTSC_0.0.1.tgz, r-oldrel: OSTSC_0.0.1.tgz


Please use the canonical form https://CRAN.R-project.org/package=OSTSC to link to this page.