Change Log

Version 0.13.0

  • Add support for Python 3.10 and 3.11, and drop support for Python 3.7.
  • Update the minimal versions required of the dependencies: * NumPy (>= 1.22.4) * SciPy (>= 1.8.1) * Scikit-Learn (>=1.2.0) * Joblib (>=1.1.1) * Numba (>=0.55.2)
  • Add an example illustrating time series clustering using pyts.transformation.BOSS transformation with different metrics (by Lucas Plagwitz).
  • Add automatic components-grouping in the Singular Spectrum Analysis for trend-seasonal decomposition with suitable example (by Lucas Plagwitz).
  • Add two new parameters in pyts.decomposition.SingularSpectrumAnalysis: chunksize allows for computing the decomposition of all the input time series using chunks (it should be a bit slower but use less memory), and n_jobs allows for running the decomposition of each chunk in parallel.
  • Set the number of initiations of K-means to compute the initial shapelets in pyts.classification.LearningShapelets: to 10 (to prevent a change of the default value in scikit-learn).
  • Replace base_estimator_ attribute with estimator_ in pyts.classification.TimeSeriesForest and pyts.classification.TSBF (to match the changes made in scikit-learn).

Version 0.12.0

Version 0.11.0

  • Add support for Python 3.8 and drop support for Python 3.5.
  • Rework the BagOfWords algorithm to match the description of the algorithm in the original paper. The former version of BagOfWords is available as WordExtractor in the pyts.bag_of_words module.
  • Update the SAXVSM classifier with the new version of BagOfWords.
  • Add the BagOfPatterns algorithm in the pyts.transformation module.
  • Add the ROCKET algorithm in the pyts.transformation module.
  • Add the LearningShapelets algorithm in the pyts.classification module.
  • Deprecated specific functions for Dynamic Time Warping (all dtw_* functions), only the main pyts.metrics.dtw() is kept.

Version 0.10.0

  • Adapt DTW functions to compare time series with different lengths (by Hicham Janati)
  • Add a precomputed_cost parameter in DTW variants that are compatible with a precomputed cost matrix, that is classical DTW and DTW with global constraint regions like Sakoe-Chiba band and Itakura parallelogram (by Hicham Janati)
  • Add a new algorithm called ShapeletTransform in the pyts.transformation module.
  • Add a new dependency, the joblib Python package, since it has been vendored from scikit-learn and it is used in ShapeletTransform.
  • [DOC] Revamp documentation in most sections:
    • User guide is much more detailed
    • A Scikit-learn compatibility page has been added to highlight the compatibility of pyts estimators with scikit-learn tools like model selection and pipelines.
    • A Reproducibility page has been added to highlight the work done in the pyts-repro repository, where we compare the performance of our implementations to the literature.
    • A Contributing guide has been added.

Version 0.9.0

  • Add datasets module with dataset loading utilities
  • Add multivariate module with utilities for multivariate time series
  • Revamp the tests using pytest.mark.parametrize
  • Add an Examples section in most of the public functions and classes
  • Require version 1.3.0 of scipy: this is required to load ARFF files with relational attributes using

Version 0.8.0

  • No more Python 2 support
  • New package required: numba
  • Updated required versions of packages
  • Modification of the API:
    • quantization module merged in approximation and removed
    • bow module renamed bag_of_words
    • Fewer acronyms used for the names of the classes: if an algorithm has a name with three words or fewer, the whole name is used.
    • More preprocessing tools in preprocessing module
    • New module metrics with metrics specific to time series
  • Improved tests using pytest tools
  • Reworked documentation
  • Updated continuous integration scripts
  • More optimized code using numba