Which of the following metrics are derived from a confusion matrix?

• A. Log Loss
• B. Mean Absolute Error
• C. ROC AUC
• D. Accuracy, Precision, Recall, and F1

Answer: (D)

A confusion matrix captures how many predictions fall into four outcomes: true positives, true negatives, false positives, and false negatives. From those four numbers, several performance measures can be calculated directly. Accuracy assesses overall correctness by taking the total correct predictions (true positives plus true negatives) over all predictions. Precision focuses on positive predictions, asking what fraction were truly positive (true positives divided by total predicted positives). Recall, or sensitivity, looks at how many actual positives were captured (true positives divided by total actual positives). F1 combines precision and recall into a single score, balancing the two by using their harmonic mean. These metrics are inherently derived from the counts in the confusion matrix, so they’re the ones that come directly from it.

Log loss, on the other hand, uses predicted probabilities for each class and the true labels, not just the discrete outcome counts from a confusion matrix. Mean absolute error is a regression metric that measures average magnitude of errors in numeric predictions, not classification counts. ROC AUC evaluates the model’s ability to discriminate across all possible thresholds and is based on true/false positive rates over varying thresholds, not a single confusion matrix.