Bootstrap AUC Confidence Interval

Problem

Bootstrap a confidence interval for the ROC AUC. Seed random, draw n_boot resamples that keep the positive and negative score lists separate, and compute a tie-aware AUC where ties count one half. Sort the replicates and return [lower, upper] at indices int(0.025 * n_boot) and int(0.975 * n_boot), clamped to the last index. Return [0.0, 0.0] when a class is missing.

Hint

Resample positives and negatives separately so every replicate keeps the original class balance.

Test cases

case 1

in: [[0,0,1,1],[0.1,0.4,0.35,0.8],200,42]

expected: [0,1]

case 2

in: [[1,0,1,0],[0.9,0.8,0.7,0.6],100,7]

expected: [0,1]

case 3

in: [[1,1,1],[0.1,0.5,0.9],50,1]

expected: [0,0]

case 4

in: [[],[],50,1]

expected: [0,0]

case 5

in: [[1,0],[0.5,0.5],100,3]

expected: [0.5,0.5]

Starter codePython

import random

def bootstrap_auc_ci(y_true, scores, n_boot, seed):
    # Returns [lower, upper]
    # Your code here
    pass

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