Class Imbalance
noun
say it: “KLAS im-BAL-unss” /ˈklæs ɪmˈbæləns/
A situation in which one category appears far more often than another in a set of data.
What is class imbalance?
Class imbalance is when one class is much rarer than another, such as 1 fraud in every 100 payments. A model can
then score 99% accuracy by never flagging anything, so you judge it by precision, recall, F1 or balanced accuracy
instead, and split with stratify.
import pandas as pd
payments = pd.Series(["ok"] * 99 + ["fraud"])
print(payments.value_counts(normalize=True).to_dict()) # {'ok': 0.99, 'fraud': 0.01}
In a sentence
- “With 1 fraud per 100 payments, class imbalance made accuracy almost useless.”
- “The shelter's data had a class imbalance: few rabbits, many dogs.”
Level: Journeyman · More AI Basics terms at this level
Updated