Training and testing

The model is trained on training data and its performance is measured on test data.

Why separate test data?

If you measure the model with the data you trained it on, you can't tell whether it has just memorised it. So the data is split in two: the model learns with the training set and its success on unseen examples is measured with the test set.

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)

test_size=0.2 sets aside 20% of the data for testing; random_state makes the split the same on every run.

The four returned parts

train_test_split returns four arrays and the order matters: first the two X parts, then the two y parts.

model.fit(X_train, y_train)          # training only
print(model.score(X_test, y_test))   # test only

Test data must never be used in training; otherwise the measured success doesn't reflect reality. This is called data leakage.

Exercise Split the data

Split the 20-example data with test_size=0.25 and random_state=0. Print the lengths of the training and test sets separated by a space.

Show example solution

from sklearn.model_selection import train_test_split

X = [[i] for i in range(20)]
y = [i % 2 for i in range(20)]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=0)
print(len(X_train), len(X_test))

Test yourself

With 100 examples and <code>test_size=0.25</code>, how many examples remain for training?
- 25
- 100
- 50
- 75
Show answer

75 — 25% (25 examples) goes to the test set and the remaining 75 are for training.

Set aside 30% of the data for testing.

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=____, random_state=0)

Show answer

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)

Match the parts with their meanings.

X_train

Training inputs
y_train

Training labels
X_test

Test inputs
y_test

Test labels

Next lesson: Feature and label notation

We Are Enhancing Your Experience

Cookie Preferences

Strictly Necessary Cookies

Analytical Cookies