Signature
Series.value_counts(normalize=False, sort=True, dropna=True)What it does
value_counts is the quickest way to audit class balance, inspect category frequencies, and decide whether a column is too rare to encode. Use normalize=True to read proportions instead of counts.
Parameters
normalizebool, optionaloptional- Return proportions instead of counts.
sortbool, optionaloptional- Order by descending count.
dropnabool, optionaloptional- Exclude missing values.
Returns
Series
Example
pythonSeries.value_counts example
import pandas as pd
labels = pd.Series(["cat", "dog", "dog", "cat", "dog"])
print(labels.value_counts(normalize=True))Output
dog 0.6 cat 0.4 Name: proportion, dtype: float64
Watch out for
- The result is sorted by count, not by category order.
- Class imbalance is easier to see with normalize=True on the training labels only.