woevidence

Module for weight of evidence transformation. More on WOE encoding here.

Getting started

Installing

git clone https://github.com/kamildar/woe_transform.git
cd woe_transform
pip install -e .

Usage

from woevidence import WoeTree
woe = WoeTree(criterion='entropy', max_depth=4, n_jobs=-1, 
              categorical_features=["education", "city"])
woe_data = woe.fit_transform(features, target)

Advantages

• missing values handling strategies, argument na_strategy allows to put NA's in special bucket or interpret them as the worst/best group
• categorical features support
• gini or entropy or custom splitting criterion
• flexible parameters for buckets (leaves): min_samles_leaf, min_samples_class, max_depth
• scikit-learn compatibility