QuestionQ74

VerifiedScaling prototypes into ML models

You are developing a binary-classification ML algorithm that determines whether an image of a classified scanned document contains a company’s logo. In the dataset, 96% of examples lack the logo, making the dataset highly skewed. Which metrics would provide the greatest confidence in the model?

Explanation

The F1 score is the harmonic mean of precision and recall, so it evaluates positive-logo detection without allowing the overwhelmingly common no-logo class to dominate the result. It gives equal importance to avoiding false positives and false negatives when neither error type is specified as more important.

Community Discussion

No comments yet. Be the first to start the discussion!