Research tasks and references for iCREDITs project
Objective 1: Selection of Feature Extraction Method
Compare the following feature extraction techniques:
(1) DFT (Discrete Fourier Transformation),
(2) DWT (Discrete Wavelet Transformation),
(3) PCA (Principal Component Analysis), and
(4) Shaplets.
(5) Any others?
References
- Feature extraction techniques from Time Series PPT
- R packages for DFT
- R packages for DWT
- Package rwt: Provides a set of functions for performing
digital signal processing. http://cran.r-project.org/web/packages/rwt/rwt.pdf
mdwt method: Computes the discrete wavelet transform y for input
signal x using the scaling filter h
- Package wavelets: This package contains functions for computing and plotting discrete wavelet transforms (DWT) and maximal overlap discrete wavelet transforms (MODWT), as well as their in- verses. Additionally, it contains functionality for computing and plotting wavelet transform filters that are used in the above decompositions as well as multiresolution analyses.
wavelets.pdf
- dwt method: Computes the discrete wavelet transform coefficients for a univariate or multivariate time series.
This method was implemented using the pyramid algorithm, with the pseudocode written in pp. 100-101 in the book.
Percival, D. B. and A. T. Walden (2000) Wavelet Methods for Time Series Analysis, Cambridge
University Press.
- plot.dwt method: Plot DWT Coeffcients
- R packages for DCT
- Package dtt: This package provides functions for 1D and
2D Discrete Cosine Transform (DCT), Discrete Sine Transform (DST)
and Discrete Hartley Transform (DHT). http://cran.r-project.org/web/packages/dtt/dtt.pdf
dtt method: Discrete Trigonometric Transforms. Set type = "dct".
- R packages for PCA
- Related research papers
Objective 2.1: Semi-supervised clustering of disturbance data
References
Objective 2.2: Pattern discovery algorithms
Objective 3.1: Classification
Objective 3.2: Localization