Text mining and content categorization

Text Clustering - Decent Empirical Way to Decide Number of Dimensions

Reply
Frequent Contributor
Posts: 115

Text Clustering - Decent Empirical Way to Decide Number of Dimensions

I have some heuristics I have developed over the years working with cluster analyses. I typically like to experiment with more and work my way back. I also like my SVD Resolution to be high.

But, what is a decent recommended approach to figuring the Maximum SVD Dimensions?

For example, in my trial run I am seeing around 5 - 7 clusters. Is there a rule to generally follow in choosing the Max No. of Dimensions? What if I go even much higher to a dozen+ clusters?

Thank you.

Ask a Question
Discussion stats
  • 0 replies
  • 247 views
  • 0 likes
  • 1 in conversation