What happened to Davies-Boudin, used to locate the optimal k in the k-means model?

tonyboy9tonyboy9 Member Posts: 113 Contributor II
I know I've been using RapidMiner AutoModel a long time when the results for k-means model now fails to produce a value for Davies-Boudin. Most writers say use 'elbow' without bothering to display where the 'elbow' number comes from. If anyone knows where this is taught in RapidMiner, please send me the link. 

I depended on Davies-Boudin in the Results of AutoModel on k-means. Continue to add to the k number in results, run again. Once Davies-Boudin increases you've added too many ks. Go back and use the prior minimal k. 

Does anyone know where Davies-Boudin went in AutoModel Results?

This screen shot from an Academy tutorial shows Davies-Bouding = 2.833.



This screen shot shows k-means Results today. Where did Davies-Boudin go?

  

Best Answer

  • ceaperezceaperez Member Posts: 517 Unicorn
    Solution Accepted
    Hi @tonyboy9
    It's true, the results view losed some interesting information about models. There is a bugfix comments around this. 
    A possible way to fix it is adding a Cluster Distance Performance operator to the model proposed for Automodel module.

    Best
Sign In or Register to comment.