Skip to content
LearnPY
scikit-learnfunction reference

train_test_split

Split arrays into training and test subsets.

Signature

train_test_split(*arrays, test_size=None, random_state=None, stratify=None)

What it does

train_test_split shuffles rows and returns two subsets for each input array. A fixed random_state makes the split reproducible. Pass stratify to preserve class proportions in classification problems.

Parameters

arrayssequence of arrays
Features and targets to split together.
test_sizefloat or int, optionaloptional
Fraction or count reserved for testing.
random_stateint, optionaloptional
Seed for a reproducible shuffle.
stratifyarray_like, optionaloptional
Labels whose proportions should be preserved.

Returns

list of arrays

Example

pythontrain_test_split example
from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42, stratify=y
)

Watch out for

  • Split before fitting any preprocessing so the test set stays unseen.
  • Pass the same random_state when you need to compare models on identical rows.