How this calculator works
Add the true positives and true negatives, then divide by the total of all four cells. Counts must be non-negative whole numbers and the total must be positive. Accuracy treats each error equally; it can look impressive when the positive class is rare. Compare it with sensitivity and precision before assessing a classifier.
Formula
Accuracy = (TP + TN) / (TP + TN + FP + FN). Error rate = (FP + FN) / total.
Worked example
With 40 true positives, 50 true negatives and 5 of each error type, 90 of 100 classifications are correct: accuracy is 90% and the error rate is 10%.
Advertisement
Frequently asked questions
What does a true positive mean?
The classifier predicts positive and the reference label is also positive. The reference labels must be defined consistently.
Can a classifier have high accuracy and miss every positive?
Yes. If only 1% of observations are positive, predicting negative for everyone gives 99% accuracy and zero sensitivity.
Does accuracy measure a numerical prediction error?
This page measures classification accuracy. For continuous predictions, error measures such as mean absolute error answer a different question.