Custom Python Evaluation Metric | Dataford Interview Questions

NTT DATA evaluates binary classification outputs from AI solutions. Implement a custom F1-score metric from scratch using only basic Python arrays and arithmetic, without importing machine learning libraries.

Given two arrays, y_true and y_pred, containing binary labels, calculate the F1-score:

F1 = 2 * TP / (2 * TP + FP + FN)

Count true positives, false positives, and false negatives during one pass through the arrays. True negatives do not affect the F1-score. If the denominator is zero, return 0.0.

Formal Specification

Input: Two arrays of integers, y_true and y_pred, with equal length. Each value is either 0 or 1.
Output: A floating-point F1-score between 0.0 and 1.0.
Do not round the result inside the function.

Constraints

1 <= len(y_true) <= 10^6
len(y_true) == len(y_pred)
Each value in both arrays is either 0 or 1
Return 0.0 when TP + FP + FN is zero

Function Signature

def binary_f1_score(y_true, y_pred):

Your question is Custom Python Evaluation Metric. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem