Model Evaluation

Use this skill for rigorously assessing model performance, comparing alternatives, and diagnosing issues.

When to use this skill

• Model training complete — need performance assessment
• Comparing multiple models/algorithms
• Diagnosing overfitting/underfitting
• Hyperparameter tuning
• Production readiness check

Evaluation workflow

1. Cross-validation strategy
   • K-fold (default for most cases)
   • Stratified K-fold (classification with imbalance)
   • TimeSeriesSplit (temporal data)
   • GroupKFold (grouped/clustered data)