Beginner

Detect Label Leakage via Temporal Check

Beginner

~12 min

code completion

Detect Label Leakage via Temporal Check

Label leakage occurs when features contain information that is only available *after* the prediction target is known. In time-series settings, a feature recorded at time $t_{f}$ should not be used to predict a label recorded at time $t_{l}$ if $t_{f} \geq t_{l}$ .

You are given:

feature_times: array of timestamps at which each feature was recorded

label_times: array of timestamps at which each label was recorded

A sample leaks if feature_times[i] >= label_times[i] — the feature was observed at or after the label event.

Return a sorted integer array of the indices of leaking samples.

Example:

feature_times = [1, 5, 3]
label_times   = [3, 4, 7]

i=0: 1 < 3 → OK

i=1: 5 >= 4 → leak

i=2: 3 < 7 → OK

Result: [1]

Your task:

Implement find_leaking_indices(feature_times, label_times) returning a sorted integer array.

Example Tests

One leak at index 1 where feature_time > label_time

Input: {"label_times":[3,4,7],"feature_times":[1,5,3]}

Expected: [1]

All features recorded before labels: no leakage

Input: {"label_times":[4,5,6],"feature_times":[1,2,3]}

Expected: []

Feature time exactly equal to label time counts as leakage

Input: {"label_times":[3,4,2],"feature_times":[3,3,3]}

Expected: [0,2]

You can read the full problem statement above. Create a free account to run code in the browser, submit solutions, and track your progress.