3
$\begingroup$

I have a data set with labels that were produced by a k-means clustering algorithm. Now there is some data (with the same data structure) from another source and I wonder what is the most sensible way to label this new, yet unseen data? I was thinking about either

  • calculating the distance to the prior k-means centroids and label the data to the the nearest centroids accordingly
  • run a new algorithm (e.g. SVM) on the new data using the old data as the training set

Unfortunately, I couldn't find anything about this particular problem. There are only a few questions about the general use of k-means as a classification model:

  • Can k-means clustering do classification?
  • How to segment new data with existing K-means model?
Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ The first your approach is the assignment. The second one is classification. In general, either could be used. Classification is discriminate-then-assign, two-stage approach. It includes "discrimination" or learning stage prior assignment stage, and the stage includes some form of modeling (with assumptions). Sometimes you want the learning stage, sometimes you don't, and sometimes you can't do it with your data or settings. $\endgroup$
    – ttnphns
    Commented Mar 6, 2019 at 11:02
  • $\begingroup$ I may recommend you to do the first of your options (assignment). Plus to this, try to gather potential outliers (points which are too distant for you from all the clusters) in a separate class or leave them unassigned. $\endgroup$
    – ttnphns
    Commented Mar 6, 2019 at 11:05

1 Answer 1

Reset to default
1
$\begingroup$

You are correct on

calculating the distance to the prior k-means centroids and label the data to the the nearest centroids accordingly

The reason run a new algorithm (e.g., SVM) will not work is because clustering is different from supervised learning that you have a label for each data point. If we have new data, we still do not have their labels. So, what we can used is just the output from the clustering, i.e., centroid.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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