Discrete Shocklet Transform (DST) and Shocklet Transform And Ranking (STAR) algorithm

A qualitative, shape-based, timescale-invariant methodology for finding shapes in sociotechnical data

Associated Paper(s):

• Published Version
• Arxiv Preprint

Installation

This repo may be installed using pip:

pip install git+https://gitlab.com/compstorylab/discrete-shocklet-transform.git

Or, if you would like to interact more with Git:

git clone https://gitlab.com/compstorylab/discrete-shocklet-transform
cd discrete-shocklet-transform
pip install -e .

Primary Features

After installation you will have access to discrete_shocklets, which has all of the tools you need in order to computer the discrete shocklet transform on arbitrary time series data. See discrete-shocklet-transform/example/example.ipynb for an application of some of the functionality found in discrete_shocklets.

In addition to the library, a command line tool is also provided that should facilitate rapid application of the STAR algorithm to arbitrary time series data. The star command line tool should be available on your path following installation. Use star -h to get an overview of the different options, and be sure to note the gotchas discussed below.

Basic DST Usage

Start with some time series data

Apply the DST

Threshold the DST to identify regions of interest

Citing this repo

If you use this code as a part of academic research then please cite one or both of the associated papers listed above. Here is the bib entry associated with the published paper.

@article{dewhurst2020shocklet,
  title={The shocklet transform: a decomposition method for the identification of local, mechanism-driven dynamics in sociotechnical time series},
  author={Dewhurst, David Rushing and Alshaabi, Thayer and Kiley, Dilan and Arnold, Michael V and Minot, Joshua R and Danforth, Christopher M and Dodds, Peter Sheridan},
  journal={EPJ Data Science},
  volume={9},
  number={1},
  pages={3},
  year={2020},
  publisher={Springer Berlin Heidelberg}
}

star gotchas

Additional arguments passed to star following the named arguments are considered to be kernel function arguments. Using the default settings, star -i example will throw the following error:

Error occurred in computation of shocklet transform of test
Error: power_cusp() missing 1 required positional argument: 'b'

The default kernel function (power_cusp) has a required argument that must be filled by the user. This can be done like so:

star -i example 3