turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Data Mining
- /
- Determine # of Text Topics

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-28-2016 10:32 AM

For Text Topic node, the default setting for # of topics is 25. I would like to know how to determine the right number. For example, can I base on the # of clusters I get from Text Cluster node to decide how many topics I should set for Text Topic node?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-28-2016 12:57 PM

Hi.

I don't think the TT node provides much in the way of guidance, but you might have a look at the HP Text Miner node (HPDM tab) if you have access to it. It gives various options for selecting the number of topics based on percentage of total variance accounted for.

Here's a snippet from Help:

"Suppose that the maximum number of SVD dimensions that you specify for the **Max SVD Dimensions** property is maxdim, and these maxdim SVD dimensions account for p% of the total variance. High resolution always generates the maximum number of SVD dimensions (maxdim). For medium resolution, the recommended number of SVD dimensions accounts for 5/6*(p% of the total variance). For low resolution, the recommended number of SVD dimensions accounts for 2/3*(p% of the total variance)."

(You could also try posting to the Text Analytics Community as they are the text mining experts)

Ray