#0078

Detect Suspicious Leakage Columns

EasyML15 min15 XP

Problem

Given feature column names and target column name, return names that look suspicious. A feature is suspicious if it equals the target, contains the target text, or contains "future", "after", or "post". Compare case-insensitively.

Why This Matters

Leakage can make a model look brilliant offline and fail instantly in production. Naming checks are not enough, but they are a useful first audit.

Function Signature

def leakage_columns(features, target):

Examples

Example 1

Inputfeatures = ["age", "churn_after_30d", "plan"], target = "churn"

Output["churn_after_30d"]

The feature appears to use future churn information.

Constraints

Preserve original feature names in the output.
Preserve input order.
This is a naming audit, not proof of leakage.

CodePython

def leakage_columns(features, target):
    """
    Input example: features = ["age", "churn_after_30d", "plan"], target = "churn"
    Return: ["churn_after_30d"]
    Return the value from the function so the tests can compare it. Do not print unless the prompt asks.
    """
    # write your code here
    return []

def leakage_columns(features, target):
    """
    Input example: features = ["age", "churn_after_30d", "plan"], target = "churn"
    Return: ["churn_after_30d"]
    Return the value from the function so the tests can compare it. Do not print unless the prompt asks.
    """
    # write your code here
    return []

Visible browser tests run here when available.

Testcases1 visible / 3 hidden categories

Target and future words

Input["age","churn_after_30d","post_score","plan"], "churn"

Expected["churn_after_30d","post_score"]

Both names look suspicious.

Hidden Test Categories

case-insensitive matchexact target columnno suspicious columns