0
$\begingroup$

I have a large dataset with a lot of features, I applied PCA on it and then run it through clustering algorithm.
I have another large dataset with the same features from different area and I want to check to which cluster each point fit sepreatly. How I do it? Do I need to apply PCA to the new dataset?

$\endgroup$

1 Answer 1

1
$\begingroup$

You will have to process your new data in the same way that you processed the original data. Regarding PCA, you will need to apply the weights from the original PCA to the new data, and not do a new PCA on the new data. Using the original weights will ensure that all data, new and old, are comparable and have been treated similarly.

Regarding clustering, how to assign new data points to 'old' clusters (in respect to their locations, cutoffs, etc) depends on the type of clustering you did, but the same rule applies as for PCA - do your cluster assignment using the information (cluster locations, cutoffs, etc) obtained from clustering on the original data.

$\endgroup$
2
  • $\begingroup$ how to assign to new clusters You probably meant "how to assign new points to 'old' (in respect to their locatiions, cutoffs, etc) cluster". Right? $\endgroup$
    – ttnphns
    Sep 22, 2020 at 13:13
  • $\begingroup$ Good catch! I've edited the response. $\endgroup$
    – KirkD_CO
    Sep 22, 2020 at 17:48

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.