Lift at K

Problem

Compute lift at k: the precision among the k highest-scored samples divided by the overall positive rate.

Sort by score descending, breaking ties by the original index. Return 0.0 if k is not in [1, n], the base rate is 0, or the input is empty.

Hint

Lift compares top-k precision against what random selection would achieve.

case 1

in: [[0,0,1,1],[0.1,0.4,0.8,0.9],2]

expected: 2

case 2

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

expected: 0

case 3

in: [[0,1,1,0],[0.9,0.8,0.2,0.1],2]

expected: 1

case 4

in: [[0,1,1,0],[0.9,0.8,0.2,0.1],5]

expected: 0

case 5

in: [[],[],1]

expected: 0

Starter codePython

def lift_at_k(y_true, scores, k):
    # Your code here
    pass

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