Inertia

Also known as · WCSS · within-cluster sum of squares

The sum of squared distances from each point to its own centroid (within-cluster sum of squares, WCSS); K-Means minimises it. It is available as km.inertia_ in scikit-learn.

When to use

Plot it against K for the Elbow Method. It can only fall as K grows, so you look for the bend rather than the minimum. With n_init, scikit-learn keeps the run with the lowest inertia.

See Lec 5.

Appears in

No references yet.