What insight does a confusion matrix provide?

• A. It shows feature importances.
• B. It provides insight into the types of errors the model makes.
• C. It measures model fairness across groups.
• D. It indicates data leakage.

Answer: (B)

A confusion matrix shows how predictions line up with actual outcomes, making the kinds of mistakes the model is making visible. It breaks down performance into four outcomes: true positives, true negatives, false positives, and false negatives. Seeing these counts at a glance lets you understand not just overall accuracy, but whether the model tends to miss positive cases or to mislabel negatives as positives. That insight is essential for deciding how to adjust thresholds, address class imbalance, or apply cost-sensitive tweaks to improve performance where it matters most.

Other ideas aren’t what the confusion matrix directly provides. It doesn’t tell you which features mattered—that’s feature importance. It isn’t a fairness metric by itself, though you could derive group-specific confusion data to study fairness. And it doesn’t diagnose data leakage, which concerns how data was collected and split. The matrix’s primary value is revealing the specific error patterns the model makes.