Added - Added mapie as a requirement. Changed - `Forecaster.proba_forecast()` now uses conformal prediction from the mapie package to create confidence intervals and is much more efficient. This will be also implemented in `MVForecaster` soon. - Took out 'CIPlusMinus' as a history key in `Forecaster` and `MVForecaster`. Fixed
0.16.3
Added - Added a naive/seasonal naive estimator to `Forecaster`. Changed - `Forecaster.add_diffed_terms` no longer supports second differencing. Fixed - Fixed the warning that gets passed for the RNN and Silverkite models when `dynamic_testing` is passed as `False`.
0.16.0
Added Changed - It is now possible to skip model testing by setting test_length = 0 in `Forecaster` and `MVForecaster`. - Got rid of LastTestSetPrediction and LastTestSetActuals columns from model summary df. - Got rid of `best_fcst` DataFrame. - Gave documentation a once-over. - Refactored code to be slightly more efficient. Fixed - The scaler in `MVForecaster` was not just being applied to the test-set inputs and was being called twice per model train.
0.15.16
Added - added `ax` argument to all forecast plotting functions. Changed Fixed - removed `tensorflow` from the list of requirements and added it and `tensorflow-macos` to the optional add-on list (31)
0.15.14
Added - added a str representation for the `SeriesTransformer` object - added `exclude_models` argument to `SeriesTransformer.Revert()` and similar functions Changed Fixed - cleaned up some documentation links
0.15.12
Added Changed - fill N/A for all forecasts with a forward fill to prevent some max-value errors, both when forecasting and reverting transformations Fixed