Logistic Regression

A probabilistic binary classifier: a linear score is passed through the Sigmoid Function to give a probability, which is classified as 1 if it exceeds a threshold (0.5 by default).

When to use

Use it when you need probabilities rather than hard labels, for example ranking members by churn risk. It assumes a linear decision boundary and is sensitive to feature scale, so scale first.

See DS Lec 3.

Appears in