Balanced Accuracy
noun
say it: “BAL-unst AK-yer-uh-see” /ˈbælənst ˈækjərəsi/
A score for a classifier that averages how well it finds each category, so rare categories count as much as common ones.
What is balanced accuracy?
Balanced accuracy is the average of the recall for each class, so a rare class counts as much as a common one. A model that always guesses the common class scores 0.5 on two classes, however rare the other is, where plain accuracy might say 0.99.
from sklearn.metrics import accuracy_score, balanced_accuracy_score
actual = [0] * 99 + [1] # 1 fraud in 100
guessed = [0] * 100 # never flags anything
print(accuracy_score(actual, guessed)) # 0.99
print(balanced_accuracy_score(actual, guessed)) # 0.5
In a sentence
- “Balanced accuracy rose when the model finally caught some fraud.”
- “The do-nothing model's balanced accuracy was 0.5, which gave it away.”
Level: Journeyman · More AI Basics terms at this level
Updated