Weight of Evidence & Information Value

The implementation of Weight of Evidence (WOE) encoding and Information Value (IV). Lots of implementation out there, yet this repo offers one using PySpark for data processing.

What you need

• Python >= 3.7.0
• PySpark >= 2.4.0

Install

• Clone this repo
• Go to the root directory of the local repo
• Run python setup.py install

Quickstart

Please check the main module for the example.

A) Prepare the data

df = <spark_dataframe>
cols_to_woe = <list_of_categorical_columns_to_encode>
label_col = <label_column>
good_label = <good_label>

B) WOE Encoding

woe = WOE_IV(df, cols_to_woe, label_col, good_label)
woe.fit()

encoded_df = woe.transform(df)

C) Information Value

ivs = woe.compute_iv()