Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

SAS Enterprise Miner SVM

Accepted Solution Solved
Reply
Occasional Contributor
Posts: 5
Accepted Solution

SAS Enterprise Miner SVM

Hello.

I am using SAS Enterprise Miner 14.1.

I want to use SVM(Support Vector Machine) Node but I can't find it.

I can see the similar node HP SVM in HPDM group, but I can't find classical SVM node.

 

Could you let me know the reason of the above problem or the way I can find the classical SVM?

 

Thanks in advance.


Accepted Solutions
Solution
‎01-18-2016 10:28 PM
Super Contributor
Posts: 336

Re: SAS Enterprise Miner SVM

I didn't play around much with the SVM node, but I am a big fan of PROC HPSVM. I think you are gonna like it.

 

This paper by Bob Rodriguez and Robert Cohen explains really well what's the deal with HPDM procedures.

http://support.sas.com/resources/papers/proceedings13/401-2013.pdf

 

In my own short summary, HPDM procs were designed to make the most out of your computing power. If you are running in a single machine, they will run multithreaded. If you are running on a big appliance (grid, blade, whatever you call it), your HPDM proc will use as many nodes and threads as are available--no syntax changes needed, only specify this mode on your libname statement.

I understand that many HPDM procs were also developed as an opportunity to revisit the algorithm, add new options, and retire some others if it would benefit your results.

 

Very interested to hear how your course comes along. Just curious, what's the final project?

View solution in original post


All Replies
Super Contributor
Posts: 336

Re: SAS Enterprise Miner SVM

[ Edited ]

Starting EM14.1 you only have HPSVM node. I think the proc for SVM node was released as experimental, so it is better if you use HPSVM node.

Do you see similar/better performance?

You need SVM mostly to migrate projects, or some other reason?

 

Thanks,

Occasional Contributor
Posts: 5

Re: SAS Enterprise Miner SVM

Thanks for the prompt and kind response.

Now I think I'd better to use HPSVM node.

 

In fact, I do not fully understand HPDM group.

I just thought that every node in HPDM group is for distributed computing environments.

Because I'm scheduled to run a kind of data mining course using e-miner in this spring, I needed to know what is HPDM group and what is main difference between HPDM and classical nodes.

In that sense, I found tree, regression, and neural network in both classical model and HPDM nodes.

But I couldn't find classical SVM node.

That is why I posted the question.

 

Thanks again for your kind reply.

 

Solution
‎01-18-2016 10:28 PM
Super Contributor
Posts: 336

Re: SAS Enterprise Miner SVM

I didn't play around much with the SVM node, but I am a big fan of PROC HPSVM. I think you are gonna like it.

 

This paper by Bob Rodriguez and Robert Cohen explains really well what's the deal with HPDM procedures.

http://support.sas.com/resources/papers/proceedings13/401-2013.pdf

 

In my own short summary, HPDM procs were designed to make the most out of your computing power. If you are running in a single machine, they will run multithreaded. If you are running on a big appliance (grid, blade, whatever you call it), your HPDM proc will use as many nodes and threads as are available--no syntax changes needed, only specify this mode on your libname statement.

I understand that many HPDM procs were also developed as an opportunity to revisit the algorithm, add new options, and retire some others if it would benefit your results.

 

Very interested to hear how your course comes along. Just curious, what's the final project?

Occasional Contributor
Posts: 5

Re: SAS Enterprise Miner SVM

First of all, I appreciate your kind comments.

According to your comments, I just decided to study HPDM procs more.

Of course, your insightful summary will be great help.

 

In my data mining course, we discuss some data mining theories and pratice sas enterprise miner.

Until 2014, I used old version (E-Miner 4.3), but I decided to use newer version (14.1) from this semister.

So I'm updating my lecture notes now.

 

The subjects of term project are open, so students can choose their own topic.

However, because my research interests are mainly on text mining, many students perform their project about text mining.

Some of the subjects are topic modeling, issue tracking, issue clustering based on users' access, and so on.

 

If you recommend any interesting topic for term project, it will be great help for me and my students. ^^

 

Thanks for your supports and efforts.

SAS Employee
Posts: 106

Re: SAS Enterprise Miner SVM

How about a topiic analyis of a favorite / interesting publicly available text? Such as a great novel, legislative activity, research abstracts, etc. 

 

For example:

 

https://communities.sas.com/t5/SAS-Communities-Library/Tip-How-to-Create-a-Thematic-Timeline/ta-p/22...

 

Hope this helps,

 

Ray

Occasional Contributor
Posts: 5

Re: SAS Enterprise Miner SVM

Thanks again for the comments.
​I've performed ​text mining with news, VOC, patents, and twitter messages.
But I haven't tried novel or legislative activity.
​I'll do these if I can acquire novel or legislative activity text.

​Thanks for your great kindness!​


##- Please type your reply above this line. Simple formatting, no
attachments. -##

--
SAS Employee
Posts: 106

Re: SAS Enterprise Miner SVM

You're welcome!


SAS Employee
Posts: 24

Re: SAS Enterprise Miner SVM

If you are an instructor you should have free access to many in-depth educational materials provided by SAS' education practice:

 

http://support.sas.com/learn/ap/prof/index.html

 

(This includes text mining materials.)

Occasional Contributor
Posts: 5

Re: SAS Enterprise Miner SVM

WOW~
Thanks for the invaluable information!!!

##- Please type your reply above this line. Simple formatting, no
attachments. -##

--
☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 9 replies
  • 2189 views
  • 0 likes
  • 4 in conversation