pytseg 0.1

Tiny toolbox for time series segmentation

Tiny toolbox for time series segmentation.

The toolbox presumes a (univariate or multivariate) time series. For example, consider the following univariate time series:

Such a time series can then be segmented into distinguishable segments using the toolbox:

All segments are marked in different colors in the plot. And, finally, these segments can be assigned labels like stationarity:

Green lines indicate stationary segments of the time series.

Installation

Install the package via pip or clone this repository. To use pip, type:

$ pip install pytseg

Usage

Documentation: https://pytseg.readthedocs.io/en/latest/.

Demo notebooks can be found in the demos/ directory of this repository.

📖 Citation

If you find this code useful in your research, please consider citing:

@misc{pytseg2022,
      author = {Raoul Heese},
      title = {pytseg},
      year = {2022},
      publisher = {GitHub},
      journal = {{GitHub} repository},
      howpublished = {\url{https://github.com/RaoulHeese/pytseg}},
     }

The implemented univariate time series segmentation closely follows:

@article{PhysRevE.69.021108,
         title = {Heuristic segmentation of a nonstationary time series},
         author = {Fukuda, Kensuke and Eugene Stanley, H. and Nunes Amaral, Lu\'{\i}s A.},
         journal = {Phys. Rev. E},
         volume = {69},
         issue = {2},
         pages = {021108},
         numpages = {12},
         year = {2004},
         month = {2},
         publisher = {American Physical Society},
         doi = {10.1103/PhysRevE.69.021108},
         url = {https://link.aps.org/doi/10.1103/PhysRevE.69.021108}
        }

There is no affiliation with the authors of this article.