CODEWITHSUNDEEP
pythoncluster-analysisk-means
Redford Dirk
Redford Dirk

Reputation: 25

Is there any study on what the optimal range will be k-values in the elbow method?

My primary issue is to somehow localize the range of k-values to perform the elbow-analysis on. At present, I can only think of a range between 2 to min(num_rows, num_columns). Is there any other optimal method that might be more suitable, especially if you have a dataset that has an extremely large number of rows and columns?

P.S., also, is there some way to automate it so that we do not have to look at the elbow-plot? At present, I'm using the KneeLocator function of the Kneed library but I wanted to consider more options.

Thank you.

Upvotes: 0

Views: 72

Answers (1)

user5597655
user5597655

Reputation:

As you might know, there is not always an "elbow". Even if there is one, that does not necessarily represent the best number of clusters.

The KneeLocator function of Kneed is probably your best bet, because it finds the K with maximum curvature.

Upvotes: 1

Related Questions