WOE Information Contribution

Problem

Compute each bin's contribution to information value: (good_share - bad_share) * ln(good_share / bad_share), clipping every share to at least 1e-6. Return one contribution per bin; if either total is 0 return a list of zeros and an empty input returns [].

Hint

Summing these per-bin contributions recovers the information value of the feature.

Test cases

case 1

in: [[30,20,50],[10,40,50]]

expected: [0.2197224577336219,0.13862943611198905,0]

case 2

in: [[50,50],[50,50]]

expected: [0,0]

case 3

in: [[],[]]

expected: []

case 4

in: [[0,10,0],[5,0,5]]

expected: [6.561168566338787,13.815496742453716,6.561168566338787]

Starter codePython

import math

def woe_contributions(good_counts, bad_counts):
    # Your code here
    pass

4 test cases · run your code against them in the editor