I have a dataset with variables with different scales as shown in the figure below. I need to group individuals together and I'm testing algorithms like Kmeans and DBScan. In all tests I'm extracting the two main components with PCA.

enter image description here

When I don't apply any transformation before PCA (neither standardization nor normalization), almost all individuals are in a single cluster. The same happens when I apply one or another transformation (standardization OR normalization).

I only get meaningful results if I apply both (standardization AND normalization).

But I'm in doubt if I'm doing the correct method here, that is, if I apply the two transformations it wouldn't be causing problems in my results.

What do you think?



Your Answer

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