Confusion Matrix Calculator

A confusion matrix organizes prediction results into a 2x2 grid comparing actual targets against model predictions. From the core counts of True Positives (TP), True Negatives (TN), False Positives (FP), and False Negatives (FN), the calculator derives vital performance equations. Accuracy is calculated as (TP + TN) divided by the total sample size. Precision measures exactness using TP / (TP + FP), while Recall (or True Positive Rate) measures completeness using TP / (TP + FN). The F1 Score computes the harmonic mean of precision and recall. Finally, the Matthews Correlation Coefficient (MCC) provides a balanced quality measure even for heavily imbalanced classes.

Imagine you build a machine learning model to diagnose a specific medical condition. Out of your test cohort, the model yields 80 True Positives, 850 True Negatives, 20 False Positives, and 50 False Negatives. First, compute Precision: 80 / (80 + 20) = 80 / 100 = 0.80 or 80%. Next, calculate Recall: 80 / (80 + 50) = 80 / 130 = 0.615 or 61.5%. To find the F1 Score, apply the harmonic mean formula: (2 * 0.80 * 0.615) / (0.80 + 0.615) = 0.984 / 1.415 = 0.695. Finally, overall Accuracy is (80 + 850) / (80 + 850 + 20 + 50) = 930 / 1000 = 93.0%. This step-by-step evaluation reveals that while your model has high overall accuracy, its ability to capture all positive cases (recall) needs improvement.

When assessing your models, avoid relying solely on accuracy, especially if your dataset suffers from severe class imbalance. Always inspect precision and recall together, as optimizing for one often degrades the other. Use the Matthews Correlation Coefficient (MCC) when dealing with skewed datasets because it provides a much more honest assessment than standard accuracy or F1 scores.