Classification accuracy, also called the accuracy rate, measures the proportion of instances a classification model predicts correctly. It is the inverse of misclassification error. section 14.6.2
Calculate accuracy from the model’s classification results:
- Accuracy = (true positives + true negatives) / total instances. section 14.6.2
- Equivalently, accuracy = 1 − (false positives + false negatives) / total instances. section 14.6.2
- Interpret accuracy alongside sensitivity, specificity, the confusion matrix, and AUC rather than treating it as the only measure of predictive quality. section 14.6.2 section 14.8
NBA postseason classification
A logistic regression predicts whether 80 teams qualify for postseason play. Its confusion matrix contains 40 true positives, 19 true negatives, 11 false positives, and 10 false negatives. The error rate is (11 + 10) / 80 = 0.2625, so accuracy is 1 − 0.2625 = 0.7375, or 73.75%.
