0.3.8

Merge pull request #12 from rrieunier/patch-1

Updated pandas and numpy requirements

0.3.6

fix some bugs (see changelog), remove a dependency and update requirements

0.3.3

Patch for updating the pandas version dep.

0.3.2

Distributing the data with the package and fix the requirements for smoother pip install

0.3.1

Modify the syntax for computing the association according to the new version of dython

0.3.0

• Fix the Boruta_py feature counts, now adds up to n_features
• Fix the boxplot colours, when only rejected and accepted (no tentative) the background color was the tentative color
• Numpy docstring style
• Implement the new lightGBM callbacks. The new lgbm version (>3.3.0) implements the early stopping using a callback rather than an argument
• Fix a bug for computing the shap importance when the estimator is lightGBM and the task is classification
• Add ranking and absolute ranking attributes for all the classes
• Fix future pandas TypeError when computing numerical values on a dataframe containing non-numerical columns
• Add housing data to the distribution
• Add "extreme" sampling methods
• Re-run the NBs

0.2.3

Update syntax to stick to the new argument names in Dython

Association collinearity filtering

Introduces collinearity filtering for a mix of categorical and continuous variables