0.4.1

Added

• Dask support ([#181]

Changed

• Changed code coverage from SonarCloud to Codecov ([#183]
• Update CONTRIBUTING.md and README.md to be consistent with README.dev.md ([#183]

0.4.0

Added

• Python 3.11 support (#175).
• Support monthly and weekly data for preprocess module (#173).

Changed

• Example data used in the tutorial notebooks is now downloaded from Zenodo (#172).

0.3.1

Monthly and weekly data now supported by the preprocessor #173.

The preprocessor supports:

• detrending
• deseasonalizing
  The preprocessor is ML friendly because it support similar (scikit-learn) syntax as used for scalers / models in the ML realm: fit() and transform().

0.3.0

Note: this is the first release designed to work with Lilio, which is a spinoff package which handles the calendar generation and resampling. s2spy will focus more specifically on (sub)seasonal forecasting.

Added

• "Label alignment" functionality for RGDR, to align labels over multiple train-test splits (#144).
• A preprocessing module, which can be used to calculate climatology/anomalies and to detrend data (#152).
• Support for specifying multiple target and precursor intervals in RGDR (#153).

Changed

• A bug in the spherical area calculation of RGDR has been fixed (#133).
• Default settings for RGDR have been removed. Users now need to fully specify their RGDR setup (#133).
• The RGDR visualization plots are now called using RGDR.preview_correlation and RGDR.preview_clusters (#106).

Removed

• Calendar, resampling, and traintest modules have been moved to a separate package named Lilio (#158).

Dev changes

• Use hatch as the project manager, and ruff as the linter (#159).
• Notebooks have been moved to the docs folder, to be included in ReadtheDocs in the future (#159).

0.2.1

s2spy is a high-level python package integrating expert knowledge and artificial intelligence to boost sub-seasonal to seasonal (S2S) forecasting. It helps you achieve trustworthy data-driven forecasts by providing end-to-end solutions to your machine learning (ML) based S2S forecasting workflow including:

• Datetime operations & data processing
• Preprocessing
• Dimensionality reduction
• Cross-validation
• Model training
• Explainable AI

Fixed

• Display of images on ReadtheDocs and PyPi (#97)

0.2.0

s2spy is a high-level python package integrating expert knowledge and artificial intelligence to boost sub-seasonal to seasonal (S2S) forecasting. It helps you achieve trustworthy data-driven forecasts by providing end-to-end solutions to your machine learning (ML) based S2S forecasting workflow including:

• Datetime operations & data processing
• Preprocessing
• Dimensionality reduction
• Cross-validation
• Model training
• Explainable AI

Added

• Improve Sphinx documentation hosted on ReadtheDocs (#32 #70)
• Support max lags and mark target period methods in time module (#40 #43)
• Add traintest splitting module for cross-validation (#37)
  • Support sklearn splitters for traintest module (#53)
  • Implement train/test splits iterator (#70)
• Add Response Guided Dimensionality Reduction (RGDR) module (#68)
  • Implement correlation map function (#49)
  • Implement dbscan for RGDR (#57)
  • Support for multiple lags in RGDR (#85)
• Update Readme (#95)

Changed

• Refactor resample methods as functions (#50)
• Refactor calendars to BaseCalendar class and subclasses (#60)

Removed

• Python 3.7 support (#65)