Scikit-learn version 0.24.0 is packed with new features for machine learning. It arrived just in time for the New Year. Letβs look at the highlights! βοΈ
Faster ways to select hyper-parameters
- ICE plots
- Histogram boosting improvements
- Forward selection for feature selection
- Fast approximation of polynomial feature expansion
- SelfTrainingClassifier for semi-supervised learning
- Mean absolute percentage error (MAPE)
OneHotEncoder
supports missing values
- OrdinalEncoder can handle new values in the test set
- Recursive feature elimination (RFE) accepts a proportion of features to retain
#python #artificial-intelligence #machine-learning #scikit-learn #deep-learning