StandardScaler

noun

say it: “STAN-derd SKAY-ler” /ˈstændərd ˈskeɪlər/

A scikit-learn tool that rescales each feature to an average of 0 and a standard deviation of 1, using values learned from the training data.

What is StandardScaler?

StandardScaler is scikit-learn's tool for standardization: fit() learns each column's mean and standard deviation from the training rows, and transform() rescales any rows with them. Fitting it on the training rows only, inside a pipeline, keeps the test rows from leaking in.

from sklearn.preprocessing import StandardScaler
train = [[10], [20], [30]]    # dollars per visit
scaler = StandardScaler().fit(train)
print(scaler.mean_)                         # [20.]
print(scaler.transform([[40]]).round(2))    # [[2.45]]: rescaled with the training mean

In a sentence

  • “Putting the StandardScaler inside the pipeline kept the test rows out of its averages.”
  • “The StandardScaler learned an average spend of $20 from the training rows.”

Level: Journeyman · More Python terms at this level

Updated