When clustering a dataset, the right number $k$ of clusters to use
is often not obvious, and choosing k automatically is a hard algorithmic
problem. In this paper we present a new algorithm for choosing k that is based
on a new statistical test for the hypothesis that a subset of data follows a
Gaussian distribution. The algorithm runs k-means with increasing k until the
test fails to reject the hypothesis that the data assigned to each k-means
center are Gaussian. We present results from experiments on synthetic and
real-world data showing that the algorithm works well, and better than a recent
method based on the BIC penalty for model complexity.
Pre-2018 CSE ID: CS2002-0716
