BookmarkSubscribeRSS Feed
Hari
Calcite | Level 5
Hi,

Im working through the examples in Getting Started with SAS EM 5.3...I used the Data partition node to split the Donor_Raw_data (http://support.sas.com/documentation/onlinedoc/miner/getstarted53.zip) in to 55% and 45% for Train and Validation role data sets (and 0% for Test) by specifying in the Properties panel and then run the node. Strangely after running the properties panel displays percentage 40, 30 and 30 for Train/Validation and Test...and even the log associated with the data partition node displays a message of 40/30/30 (and same for the counts in output window)...(The target variable is Target_B whose measurement level is Binary)

I tried the same with another data file HMEQ within SAMPSIO library (by creating a new diagram) and even in that, any change in the default percentages and running the node leads to producing results as per 40/30/30 split only...

Why is this happening...

I tried searching the support web-site and see that this issue was reported for version 4.1 more than 9 years back - Problem Note 3570: Data Partition node does not work correctly with stratification - http://support.sas.com/kb/3/570.html But again this was resolved in 4.2 version

I also see one issue from 3 weeks back- Problem Note 30664: Model Comparison node might produce incorrect choice when data is partitioned - But this is not for the data partition output itself but effect on subsequent node...

Sincerely,
Hari
1 REPLY 1
DougWielenga
SAS Employee

We have not seen this specific issue reported, but there are situations that can persist that might lead to unexpected behavior including the following:

 

1 - Are their problems with that specific instance of the node or with the diagram itself?   Diagrams depend on a great many files as do the individual nodes.  Should one of those files encounter a problem (e.g. corrupted, written on a bad hard drive sector, etc...), you can see strange behaviors.  Solution - In cases like this, replace the node with a new one and see if the behavior persists.  At times it might be necessary to build a new diagram (or even a new project) but these latter cases are extremely rare.  

 

2 - Has the project been imported from a prior version or different operating system?  Importing projects from older versions of the software or from different operating systems might lead to problems.  Solution - In cases like this, it might be necessary to build a new diagram (or even a new project).

 

3 - Are you using the version of Java that was tested for your installation?   Java versions are not always backwards compatible but your Java installation often prompts the user to install updates.  Updating to newer versions of Java that have not been tested with your version of SAS Enterprise Miner can lead to strange behavior.    Solution - In cases like this, you should modify your configuration so that the correct version of Java is being used by SAS Enterprise Miner.  

 

4 - Are you having any issues with memory -- either Java memory, disk space, or RAM?   When Java runs out of memory, it often leads to strange unreplicatable behavior.  Likewise, if you are running low on disk space or your memory is being heavily used since there are several other applications running, you can also encounter unusual behavior.  Solution - In cases like this, you should make more memory available, either through modifying your Java memory allocation, shutting down other applications, or archiving files to make more disk space available to SAS Enterprise Miner.  

 

I hope this helps!

Doug

 

sas-innovate-2024.png

Don't miss out on SAS Innovate - Register now for the FREE Livestream!

Can't make it to Vegas? No problem! Watch our general sessions LIVE or on-demand starting April 17th. Hear from SAS execs, best-selling author Adam Grant, Hot Ones host Sean Evans, top tech journalist Kara Swisher, AI expert Cassie Kozyrkov, and the mind-blowing dance crew iLuminate! Plus, get access to over 20 breakout sessions.

 

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
  • 1 reply
  • 874 views
  • 0 likes
  • 2 in conversation