So with numerical data, an algorithm finds the best spots for the seeds based on the scatter plot and calculates wcss on it. But when using the categorical data, does it have to calculate anything? I mean isn't seeding useless here? Algorithm just clusters by their category directly right?

Agreed, I would also like to know. What is the difference between "clustering" on a categorical vs. using `data.group_by("Language")`?

Follow this topic