Options
Clustering by variable
Newbie
Hi everyone!
I'm working on a group project with Rapidminer and my classmates and I are trying to divide our data into some clusters, but we don't know how to chose the variable to do the clustering since it seems like Rapidminer automatically uses the one of the first column of the dataset we use.
We wanted to define them by frequency but in the screenshots you can see the results we actually got.
Can anyone please help us sort out how to proceed if for instance we want to create these clusters by frequency?
I'm working on a group project with Rapidminer and my classmates and I are trying to divide our data into some clusters, but we don't know how to chose the variable to do the clustering since it seems like Rapidminer automatically uses the one of the first column of the dataset we use.
We wanted to define them by frequency but in the screenshots you can see the results we actually got.
Can anyone please help us sort out how to proceed if for instance we want to create these clusters by frequency?
0
Best Answer
-
Options
BalazsBarany
Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert Posts: 955 
Unicorn
Hi!
RapidMiner uses all attributes with most clustering algorithms, e. g. k-Means.
It's a good idea to remove the ID from processing by the clustering operator by using Set Role and setting its role to "id". That way it won't be considered for the distances that determine the clustering.
For k-Means and other distance based algorithms it's a good idea to use Normalize if you have numeric attributes on different scales. Otherwise, the attribute with the largest values will dominate the distance.
Regards,
Balázs1
