BookmarkSubscribeRSS Feed
singas
SAS Employee
The score node in E-miner (version 14.2) does not throw error if required input scoring variables are not present in scoring data. I came across this issue while working on customer project. By mistake, Some of the variables in scoring data were named differently than input training data. If we would have been scoring outside E-miner, the sas code would throw error saying "variable not found". But, in case of score node in E-miner, it did not throw error. It ran fine with green check on. The results were misleading since all probabilities/scores were under predicted(or over predicted). I would propose that the score node must fail if required input scoring variables are not present in scoring data. This requirement is as fundamental as the need to define target variable before building model. Please share your ideas. In order to replicate the issue, I have used two dummy sample data available in permanent libraries of Eminer: 1. samsio.HMEQ is used as training data 2. sashelp.CLASS is used as scoring data Both these data are not remotely connected. They don't have a single common variable. Even then, the score node runs fine without error. Please see attachment for E-miner diagram. On further investigation, it was found that score node treats absent variables as variable present with all the values missing.
Untitled.png
3 REPLIES 3
Reeza
Super User

You can post ideas in the Ballotware though this is one I'd probably talk to SAS Support about as well.


I'd expect an error or an option to override that manually if required, but not as the default value.

 

Idea/suggestions go here and users can vote on them:

https://communities.sas.com/t5/SASware-Ballot-Ideas/idb-p/sas_ideas

 

singas
SAS Employee

Reeza,

 

Thanks for the link. The issue was also taken up with tech support. However, seems like this is how score node is built.

Reeza
Super User

Then I'd definitely add it to the ballot so people can vote to change it. 

 

There are always 'defaults' in programs that people can disagree on. 

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

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
  • 867 views
  • 3 likes
  • 2 in conversation