Nixtla/mlforecast v0.5.0

on GitHub

Breaking changes

• remove dashes from feature names by @jmoralez in #69
• replace predict_fn with callbacks by @jmoralez in #73

Features

• add MLForecast.from_cv by @jmoralez in #71
• allow models to be dict by @jmoralez in #72
• Add step size argument to cross validation method by @FedericoGarza in #74
• Add new_data argument to predict method (allow transferability) by @FedericoGarza in #79
• Perform cross validation without refitting the models by @FedericoGarza in #81
• Support one model per horizon approach by @jmoralez in #80
• support multiple models in cross_validation by @jmoralez in #84

Bug fixes

• Remove dynamic_dfs argument from cross_validation method by @FedericoGarza in #82

Documentation

• add getting started docs section by @jmoralez in #64
• Add cross-validation tutorial by @FedericoGarza in #76
• Add electricity peak forecasting tutorial by @FedericoGarza in #77
• Improve description preprocessing ERCOT dataset by @FedericoGarza in #78

Maintenance

• set methods on GroupedArray and preserve id_col in TimeSeries.fit_transform by @jmoralez in #70