STOCKOUT THRESHOLD SWEEP
Beginner
Difficulty: Easy Tags: Classification, Metrics & Evaluation
Story
Walmart's stockout classifier ships with a default 0.5 decision threshold, but the business cost of a missed stockout versus a false alarm isn't symmetric, so the team wants the threshold that actually maximizes F1 on their validation set, not the framework default.
The Math
For each distinct score value t present in the data, using t as the decision threshold (predict
positive if score >= t), compute F1(t); return the t maximizing F1 (ties broken by the
larger threshold).
Input Format
n
score_1 y_1
...
score_n y_n
Output Format
best_threshold best_f1, 6 decimals.
Constraints
1 <= n <= 10^5- Time limit: 1.0 second.
Example
Input
8
0.9 1
0.8 1
0.75 0
0.6 1
0.55 0
0.4 0
0.3 1
0.2 0
Output
0.600000 0.750000
Keywords
classificationmetrics-and-evaluation
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.