Options

Find the optimal number of K in clustering

TahminehTahmineh Member Posts: 1 Newbie
Hi everyone. 
i want to determine the optimal number of k in clustering. i applied silhouette criteria. but all of my Avg. within centroid distance according to different numbers of k are negative. 
how can i choose the best one between this negative numbers.

Answers

  • Options
    Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn
    Did you try the x-means operator?  This does not use the silhouette method but instead uses AIC/BIC penalties for adding additional clusters.  It is usually a good operator to check if you don't have a strong a priori preference for your k value.
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
Sign In or Register to comment.