BookmarkSubscribeRSS Feed
ycenycute
Obsidian | Level 7

In the property panel for the clustering node, we have support percentage, is this the minimum support percentage for selecting association rule? If yes, why in the results, we see rules below this threshold. Say if we set the support percentage as 5, in the results, we see rules below 5, some having 4.**. Do I misinterpret the meaning of support percentage? 

 

 

3 REPLIES 3
sbxkoenk
SAS Super FREQ

Minimal support percentage in clustering node??

I think you are mixing up things.

"Support" looks to me like a metric of association rule analysis (market basket analysis).

Koen

ycenycute
Obsidian | Level 7
In the property panel of the association node in SAS EM, we have one parameter called support percentage, and here is the explanation of this parameter:

"Specifies the minimum transaction frequency to support an association. The frequency is expressed as a frequency. The frequency is expressed as a percentage."
sbxkoenk
SAS Super FREQ

Hello @ycenycute ,

 

OK. So you are using the association node and NOT the clustering node (as the title is suggesting).

 

This might be an explanation for what you see :

The Support Percentage refers to the "proportion of the largest single item frequency, and not the end support".  Suppose you had 500 transactions, and the item which occurs in the highest proportion of transactions occurs in 300 of the transactions.  Suppose also that you set the Minimum Support Percentage to 5%.   The most frequently occurring item occurs in 300 transactions, and 5% of 300 is 15.  You would expect to see rules that occur in at least 15 of the 500 transactions which is about 3% of the time overall but still represents at least 5% of the largest single item frequency.  Setting in terms of the overall transactions would often generate way too many or too few rules, but connecting it to the largest single item frequency makes the results more useful by default.  You can always change it larger or smaller to impact how many rules are generated.

 

Let me know if you think this is contradictory to what is said in the doc :
SAS® Enterprise Miner™ 15.1: Reference Help
Association Node
https://go.documentation.sas.com/doc/en/emref/15.1/p02nz0ihv4ir5gn1p8meajwvxz8j.htm

 

Kind regards,

Koen

Ready to join fellow brilliant minds for the SAS Hackathon?

Build your skills. Make connections. Enjoy creative freedom. Maybe change the world. Registration is now open through August 30th. Visit the SAS Hackathon homepage.

Register today!
How to choose a machine learning algorithm

Use this tutorial as a handy guide to weigh the pros and cons of these commonly used machine learning algorithms.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 3 replies
  • 498 views
  • 0 likes
  • 2 in conversation