Threshold for Given Precision

Problem

Find the threshold with the highest recall subject to precision >= target. Sweep unique probabilities from high to low.

Return the threshold (0.0 if none qualifies), with ties keeping the higher threshold.

Hint

Lower thresholds raise recall but erode precision; pick the most permissive feasible one.

case 1

in: [[0,0,1,1],[0.1,0.4,0.35,0.8],0.8]

expected: 0.8

case 2

in: [[0,0,1,1],[0.1,0.4,0.35,0.8],0.6]

expected: 0.35

case 3

in: [[0,0,1,1],[0.1,0.4,0.35,0.8],0.5]

expected: 0.35

case 4

in: [[0,0],[0.1,0.9],0.5]

expected: 0

Starter codePython

def threshold_for_precision(y_true, probs, target):
    # Your code here
    pass

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