mix two clusters ?


#1

Hi there,

I’m wondering how can I change the attributs of two clusters so that they are not clearly separate from each other? (e.g. for a dataset with 2 clusters and 100 attributes (include nominal and numerical attributes))
Thanks.

N.


#2

How do you assign cluster membership? Usually clustering algorithms try to find clusters that are similar within the clusters and dissimilar between the different clusters, i.e. more less cleary seperate the records from each other. I’m wondering what is the point in assigning clusters that do not seperate the data?


#3

Hallo,

Thanks for the reply…I want to use different artificial data sets with two classes to test if my model works…so I’m trying to test with data sets that:

  • have perfectly separated 2 classes

  • have 2 classes that are not equally assigned to objects

  • have 2 classes that are somehow correlated with each other

That’s why I’m wondering the way to mix the data points of two classes…

Best Regards,
N.


#4

I’m sorry but I still didn’t get what you try do achieve. Can you elaborate a little more what “perfectly seperated classes”, “classes not equally assigned to objects” and “somehow correlated classes” mean? And what kind of model did you train and want to test now? Can you share your workflow as well?


#5

You might want to checkout the fuzzy clustering.
This is the node:

Cheers, Iris