Inertia
noun
say it: “ih-NUR-shuh” /ɪˈnɝʃə/
In k-means, the total of the squared distances from each example to its group's center, used to measure how tight the groups are.
What is inertia in k-means?
Inertia adds up each example's squared distance to its cluster's center: the lower it is, the tighter the
clusters. It always drops as you add clusters, reaching 0 when every example is its own cluster, so you look for where
it stops dropping much. scikit-learn's KMeans keeps it in inertia_.
from sklearn.cluster import KMeans
visits = [[1, 30], [2, 28], [1, 32], [6, 5], [7, 6], [6, 4]] # visits a week, dollars a visit
for k in (1, 2, 3):
print(k, round(KMeans(n_clusters=k, random_state=0).fit(visits).inertia_, 1))
# 1 986.3
# 2 11.3 a big drop: two real groups
# 3 4.7
In a sentence
- “Lower inertia alone can't choose k, because it always drops as k grows.”
- “The inertia fell sharply from one cluster to two, then barely moved.”
Level: Freelancer · More AI Basics terms at this level
Updated