mfrdixon/OSTSC: Over Sampling for Time Series Classification

Oversampling of imbalanced univariate time series classification data using integrated ESPO and ADASYN methods. 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. EPSO estimates the covariance structure of the minority-class samples and applies a spectral filer to reduce noise. ADASYN is a nearest neighbor interpolation approach which is subsequently applied to the EPSO samples. This code is ported from a matlab implementation by Cao et al. [2011] and adapted for use with Recurrent Neural Networks implemented in 'TensorFlow'.

Package details

AuthorMatthew Dixon [ctb, cre], Diego Klabjan [ctb], Lan Wei [aut, trl]
MaintainerMatthew Dixon <>
Package repositoryView on GitHub
Installation Install the latest version of this package by entering the following in R:
mfrdixon/OSTSC documentation built on May 18, 2019, 8:13 p.m.