Skip to content

0.10.0 - 2022-02-04

base

  • Introduce base.MiniBatchTransformer. Add support for mini-batches to compose.TransformerUnion, compose.Select, and preprocessing.OneHotEncoder.

checks

  • Created this module to store estimator unit testing, rather than having it in the utils module.

compose

  • Split compose.Renamer into compose.Prefixer and compose.Suffixer that respectively prepend and append a string to the features' name.
  • Changed compose.Renamer to allow feature renaming following a mapping.

evaluate

  • Refactored evaluate.progressive_validation to work with base.AnomalyDetectors.

facto

  • Added debug_one method to BaseFM.

feature_extraction

  • Make the by parameter in feature_extraction.Agg and feature_extraction.TargetAgg to be optional, allowing to calculate aggregates over the whole data.
  • Removed feature_extraction.Lagger and feature_extraction.TargetLagger. Their functionality can be reproduced by combining feature_extraction.Agg and stats.Shift.
  • feature_extraction.Agg and feature_extraction.Target now have a state property. It returns a pandas.Series representing the current aggregates values within each group.

metrics

  • metrics.ROCAUC works with base.AnomalyDetectorss.

misc

  • Created this module to store some stuff that was in the utils module but wasn't necessarily shared between modules.
  • Implement misc.CovMatrix.

reco

  • Renamed the Recommender base class into Ranker.
  • Added a rank method to each recommender.
  • Removed reco.SurpriseWrapper as it wasn't really useful.
  • Added an is_contextual property to each ranker to indicate if a model makes use of contextual features or not.

stats

  • stats.Mean, stats.Var, and stats.Cov each now have an update_many method which accepts numpy arrays.

utils

  • Removed utils.Window and use collections.deque instead where necessary.