If we've built a model, the next question is simple—"Does this model really perform well?" Yesterday we saw the pitfall where simple accuracy on imbalanced data misleads us about a model's quality. Today, moving beyond that pitfall, we'll learn three basic evaluation metrics (accuracy, precision, recall) as easily as possible, and address two typical symptoms of wrong learning (overfitting and underfitting).
AIF-C01 doesn't expect you to memorize complex formulas for these metrics. Instead, it asks for intuition about "what does precision care about, and what does recall care about?"