- Update failing param grid for exponential smoothing models
- Update depencencies
- Silently skip fialing autosarimaxin large scale CV
- Fix issues with unstable models picked during CV
- Improve logging (hcb_verbose flags)
- Update docs
- Update depencencies
- Add random states to examples and model grid configuration
- Fix damped_trend in statsmodels model definition
- Add support for python 3.6
- Fix loading of country code column for model selector
- Improve compatibility with sktime
- Add before_days, after_days, bridge_days features to the holiday transformer for better modeling of after, before and between holiday effects
- Add Theta model
- Add support for simultaneous usage of multiple holidays
- Add progress bars for model selection
- Improve plotting
- Column names created by Seasonality and Holiday transformers start with _ to prevent name clashes
- Fix country column handling in ModelSelector
- Improved experience of trying examples through Binder (pre-cached docker image)
- Adding templates for GitHub issues
- More informative error messages when trying to use wrappers without installed dependent packages
- Updating examples and dependencies to enable wider usage with core installation
- Adding Windows distribution
- Other internal fixes (CI, CD, conda distribution, pre-commit, code cleanup, test coverage)
- Python packaging fixes
- A time-series forecasting library developed by the data science team of Heidelbergcement using a scikit-learn compatible API and popular ML libraries as backend
- First public release