Added - Added the `RobustScaler` transformer and added it to the default optimal transformation search. - Added `'robust'` as a valid normalizer argument when forecasting with scikit-learn esitmators. Changed Fixed
0.18.8
Added Changed - Made an explicit error message when the RNN model does not have enough observations to evaluate (58). - Changed the title of the loss plot from the RNN model. Fixed - Fixed how the prophet model creates its externals regressors dataframe to avoid a `ValueError` (55). - Fixed an error with the RNN model when forecast horizon is 1.
0.18.7
Added Changed - Changed requirements to avoid a dask/lightgbm error when importing the Forecaster module (46). Fixed - Calling `auto_forecast()` without tuning a model first no longer raises an error when banking the model's history (52). - RNN and LSTM models no longer raise errors with default CV parameters.
0.18.6
Added - Added `exclude` argument to `Forecaster.plot()`, `Forecaster.plot_test_set()`, and `Forecaster.plot_fitted()`. Changed Fixed
0.18.5
Added Changed - The combo model can now accept one-model arguments to facilitate auto-selecting "best" models. Fixed - Updated requirement version of lightgbm (46)
0.18.4
Added Changed - Took out `infer_datetime_format` args from pandas functions since they are deprecated. - Modified example grids. Fixed