Uplift Gain at Fraction

Problem

Compute the uplift gain of the top fraction of a ranked list. Sort by predicted uplift descending (ties by index), take the top k = max(1, int(n * fraction)) samples, and subtract the overall treated-minus-control response difference from the top-k difference; a treatment group missing from the top k contributes a 0.0 rate. Return 0.0 for invalid fractions or a missing treatment group.

Hint

Gain measures how much better the top-ranked segment responds compared with treating everyone at random.

Test cases

case 1

in: [[1,0,1,0],[1,1,0,1],[0.9,0.8,0.7,0.6],0.5]

expected: 0.5

case 2

in: [[1,0,1,0],[1,1,0,1],[0.1,0.2,0.3,0.4],0.5]

expected: -0.5

case 3

in: [[1,1,0,0],[1,0,1,0],[0.9,0.1,0.8,0.2],0.5]

expected: 0

case 4

in: [[1,1,0,0],[1,0,1,0],[0.9,0.1,0.8,0.2],0]

expected: 0

case 5

in: [[],[],[],0.5]

expected: 0

Starter codePython

def uplift_gain_at_fraction(treatment, outcome, uplift_scores, fraction):
    # Your code here
    pass

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