This release includes some features and bugfixes.
New features
- possibility to provide a custom user-defined scoring function to `Scoring.compute_loss()` and dependencies (`EndaEstimator.get_loss_training()` and `Backtesting.backtest()`)
- add possibility to perform a backtesting with an already-trained model
- add a custom logger
- add a verbose argument to `Holidays` functions and `Backtesting.backtest`
- enable upsampling for one-row dataframe
Bug fixes
- fix `Resample.forward_fill_last_record()` when all columns are of dtype object (and thus fix `Resample.upsample_and_interpolate()`)
- make forward-fill upsampling dtypes-consistent
- require h2o module to be >=3.38.0.1