0
$\begingroup$

I am doing an assignment on cluster analysis. My settings in SAS enterprise Miner are Property settings

I am getting one cluster with only one observation in it. Why is it so? how do I interpret this?

enter image description here

$\endgroup$
1
  • $\begingroup$ If I were you I'd reduce the number of clusers, it's best to visualise the data first, if possible, so you have some idea of the number of clusters required. $\endgroup$ – Morgan Ball Dec 9 '16 at 12:26
1
$\begingroup$

As far as I can interpret from your settings, you seem to be using K-Means clustering. The thing with K-means is that your centroids are initialized randomly. (At least the implementation used in Enterprise miner. K-means is computationally very intensive, so this is done to save time.) This one data point seems to be pretty far off from the centroids and therefore does not fit in with the other clusters. If you lower the maximum number of clusters, then this point will probably fit in with another cluster eventually. The problem with that, is that you are possibly imposing a structure on your data that is not there. What always helps is visualizing your clusters in some sort of scatterplot. Good luck!

$\endgroup$
1
$\begingroup$

You are getting an outlier. Cluster analysis can be used for detecting outliers as you can see.

Check the Transformed balance variable, the centroid for that cluster is 0 which is very different from the rest of the clusters' centroids which are between 8 and 9.

Ask yourself why that observation is getting those specific values, and in case it is an error, remove it from the sample of the analysis and perform it again.

Good Luck!

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.