We have datasets that have a Gaussian distribution.
,Data were obtained from different, irregular, and multimodal Gaussian distributions
How can we use the k-means clustering method for highly optimal clustering so that the most statistically similar data are in the same group?
Recent researches on subspace clustering have shown that applying k means directly to original data can not guarantee an optimal solution(Noise is a factor). You can look up subspace clustering where k means is usually applied to a low dimensional embedding matrix to obtain a clustering matrix.
Have a look at Article Global K-Means (GKM) Clustering Algorithm: A Survey
Article Rethinking k-means clustering in the age of massive datasets...
and
If the dataset is known to have Gaussian distributions, then I suggest instead of k-means you can try GMM based clustering. An example using MATLAB is shown here : https://in.mathworks.com/help/stats/clustering-using-gaussian-mixture-models.html
- Badges
- Science topic