I have done Kmeans clustering on my data based on three main features. The main scatter plot is as below:

enter image description here

But when Kmeans clusters the data it seems that a part of data points are being flattened. I have put the clustering result from another viewpoint which seems to be weird:

Clustering Results by Kmeans Method

As seen the red cluster is flattened? I have no idea about the reason!! Has anyone encountered with similar situation ever?

  • $\begingroup$ Are you sure your plotting code is correct? Also, this data set should cluster much better with DBSCAN. $\endgroup$ – Anony-Mousse Aug 14 '17 at 19:15
  • $\begingroup$ Why do you think that DBSCAN is a better choice? I run the Kmeans code from Matlab function set, As far as I know the code is ok./I have written the code below: %%%%%%%% Kmeans code to obtain 3 clusters%%%%%%%%% || size Feats || idx = kmeans(Feats,3,'MaxIter',10000,'Display','Iter','Replicates',201); $\endgroup$ – Isa Emami Aug 15 '17 at 6:13
  • 1
    $\begingroup$ The data had arbitrarily shaped, density clusters. Not gaussians of the same size. Therefore, try DBSCAN. And since the coordinates changed, I assume your visualization code is broken. $\endgroup$ – Anony-Mousse Aug 15 '17 at 7:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.