Interview OS
Back to AI / ML

Library

Evaluation metrics

Precision, recall, F1, AUC.

Eval

Overview

Choosing the right evaluation metric decides whether a model is actually good. Accuracy misleads on imbalanced data, so you weigh precision, recall, F1, ROC-AUC, and business cost against the problem's real objective.

How it works

Eval
PredictMatrixScoreOutputOutputsŷ vs yConfusioncmP/R/F1/AUCimbalance→F1Decision
ClientServiceEdgeData

Step by step, with examples

  1. 1

    Outputs

    • Compare predictions to labels.
  2. 2

    Confusion

    • Count TP/FP/FN/TN.
  3. 3

    P/R/F1/AUC

    • Choose the metric that fits the problem.
  4. 4

    Decision

    • Set the threshold for the trade-off.
    • Example: PR vs ROC

Overview

Pick metrics that match the cost of errors: precision/recall for imbalance, AUC for ranking, RMSE/MAE for regression.

Key idea

Accuracy lies on imbalanced data — always state the base rate.

Common pitfalls

  • Optimizing accuracy on skewed classes
  • Ignoring calibration

Where this content comes from

For full transparency, this content is curated and verified from these sources:

Peer-reviewed ML papers & model cardsProduction ML engineering referencesOppZen-authored ML eval rubrics