What is a confusion matrix?

• A. A table used to assess the performance of a classification model on a test dataset where the actual values are known.
• B. A graph showing ROC curve.
• C. A table listing hyperparameters.
• D. A chart of training loss over time.

Answer: (A)

A confusion matrix is a table that shows how a classification model’s predictions compare to the true labels on a labeled dataset. It records actual classes versus predicted classes, with counts in each cell. In a binary case, it’s a 2x2 grid containing true positives, true negatives, false positives, and false negatives, which lets you see not just overall accuracy but exactly what kinds of mistakes the model makes. In multiclass problems, it expands to a k by k matrix that reveals which classes are being confused with which others. This requires known actual values to evaluate the predictions, so you can compute metrics like accuracy, precision, recall, and F1, and diagnose where the model’s performance is weak. The other options describe different tools (a ROC curve, a list of hyperparameters, or a training loss chart) that don’t capture the detailed correspondence between predicted and actual labels.