What is the “seed” in Weka's SimpleKMeans clus

问题:

I'm using Weka's SimpleKMeans clusterer on a set of data.

But I'm unsure what the seed value is, what it does or how it affects the data.

i.e. How does a higher or lower seed value affect the result as oppose to the default value of 10?

回答1:

Seed is just a random numbers seed. Once seed is fixed, even a randomized algorithm behaves deterministically. KMeans is not deterministic, so if you want repeatable results - you fix a seed. However there is completely no relation between exact value of the seed and the results of KMeans clustering.