Data Mining (77 page)

Within-cluster variations, after initial random distribution of samples, are

And the total square-error is

When we reassign all samples, depending on a minimum distance from centroids M
₁
and M
₂
, the new redistribution of samples inside clusters will be

New clusters C
₁
= {x
₁
, x
₂
, x
₃
} and C
₂
= {x
₄
, x
₅
} have new centroids

The corresponding within-cluster variations and the total square-error are