jwexler Tracker

NEW Automation and Interpretability - Visual Data Mining and Machine Learning 8.5

jwexler — Wed, 27 Nov 2019 17:04:43 GMT

Happy Holidays everyone! I wanted to bring everyone up to speed on the latest features and innovations released last week in VDMML 8.5. This was part of a larger release of Viya 3.5. In this release, you can now run automated feature engineering and modeling with the click of a button inside Model Studio, or embed ' a data scientist in a box' using the new machine learning pipeline automation api. This API is fully documented with examples on developer.sas.com. You can catch the latest new features in action here in my Snapshot Video. Read on to hear about these new features.

You would be hard pressed to attend a conference, read a whitepaper or blog, or attend a webinar without hearing about automated machine learning (Auto-ML). You do not need to be a data scientist to take advantage of the latest methods in machine learning to identify fraud, predict churn, or classify credit-worthiness. The latest release of VDMML 8.5 provides one-click access to dynamically generate a pipeline for you. This dynamic process will run hundreds of techniques for you to identify potential data quality issues, candidate transformations, and the right machine learning techniques to address your business problem. Tjhe system will keep you updated as it is running your process.

Once the process completes you will be provided with a pipeline that represents the top features and models that were identified. The beauty of this process is that it is not black box. You have full transparency on the results, and can edit or add in additional techniques. Analyze the most important attributes using Kernal Shap and gain an understanding of how to interpret it using natural language.

We also released 2 new nodes to support automated feature engineering (Feature Machine) and automated modeling (Model Composer). These nodes allow you complete control over the transformation policies and machine learning models run. Model Composer allows you to run autotuning across multiple algorithms, allowing the processing to take advantage of your compute infrastructure - as the processing runs multiple rounds, priority is given to the most predictive algorithms.

In summary, I am extremely excited to share these new features with everyone. There is so much more to come in this area from SAS, stay tuned more powerful features in upcoming releases.

Jonathan Wexler

Advisory Product Manager - SAS AI and ML

Re: Segment Profile is not returning the 12 clusters provided by HP Cluster

jwexler — Thu, 23 Aug 2018 19:37:26 GMT

That's great that it's working now. It looks like it's data dependent, as I suspected. Somehow your obersvations are being scored/put into 1 cluster.

Best advice for a quick solution is to open up a quick tech support track to get hands-on advice from our trained staff.

support.sas.com

Good luck!

Re: Segment Profile is not returning the 12 clusters provided by HP Cluster

jwexler — Thu, 23 Aug 2018 18:24:17 GMT

That should not matter for what you are doing. Can you try creating your HP Cluster with a max of 2 clusters? Lets see if your problem is data-dependent.

Re: Segment Profile is not returning the 12 clusters provided by HP Cluster

jwexler — Thu, 23 Aug 2018 17:23:46 GMT

Hi, can you send over the properties panel screenshot for your HP Cluster node? I tried your flow using the sample HMEQ dataset that is included with EM, and segment profile worked as expected. I also wonder if your segment sizes are too small for segment profile, because there may be a cutoff value.

Thanks

Jonathan

Re: Save Proc Neural parameter estimates in a file and re-use them to make predictions

jwexler — Thu, 23 Aug 2018 17:06:38 GMT

Hi, have you tried using data step scorecode (code statement)? Proc neural produces scorecode so you can use this scorecode in a scoring process with data step.

Re: Hide Enterprise Miner custom tab

jwexler — Thu, 10 Aug 2017 14:33:15 GMT

HI Gabee. There is no SAS-supported way to hide tabs.

Good luck!

Jonathan

Re: local installation of SAS miner and connection to libraries on a SAS server.

jwexler — Thu, 10 Aug 2017 14:31:43 GMT

Hi Eran! Depending on your setup (whether you have EM Server or EM Desktop), you can map or assign libraries in a few different places within Enterprise Miner. See this section in your Enterprise Miner Documentation:

Allocating Libraries for SAS Enterprise Miner

In order to map libraries, your EM machine needs to be properly licensed and have visbilibikity and proper access to the locations you want to map.

Good luck!

Jonathan

Machine Learning using SAS code, Python and Visual Analytics -- all in one place! Viya!

jwexler — Fri, 24 Mar 2017 14:52:19 GMT

Good afternoon everyone. I’m very excited to share details about our updated version of SAS Visual Data Mining and Machine Learning on SAS Viya.

Before I dive into the details, it’s important to understand the thinking behind this revolutionary release. As the need for machine learning as skyrocketed, so has the need to access methods from multiple entry points. Organizations are often made up of a diverse set of individuals with varying backgrounds in computer science, statistics, machine learning, and business. Accompanying these backgrounds is the myriad of analytical ‘tools’ that you need to solve modern business problems. Examples may include SAS, which includes the SAS language and our graphical users interfaces. You may have a background in Python, R, Java or Lua. You may even be an application developer who wants to build applications from the ground-up using APIs. No matter your skillset, you should be able to use your language and interface of choice.

Welcome SAS Visual Data Mining and Machine Learning on SAS Viya! At its core, SAS Viya is built upon a common in-memory analytic framework, using ‘actions’. These actions are atomic analytic activities, such as selecting variables, building models, generating results, and outputting score code. The actions or components are modular by design. We have exposed these actions via SAS Procedures, Python, Java, Lua and RESTful APIs (R will be released soon). No matter the language or interrace, you WILL get the same answers for the same actions, whether you use a procedure or a python call into SAS Viya. Start your analysis in SAS, and then continue it in Python, all using the same in-memory data – no duplication.

We have also exposed these analytic actions with the Visual Analytics framework. Now you can build a two-layer Neural Network using LBFGS and compare it to a Gradient Boosting model with 500 trees, all within an interactive environment.

There’s so much more to this offering. We’ve enhanced our capabilities in autotuning so that you intelligently search the hyperparameter space for the best combination of values that addresses the model objective – that is, misclassification, Lift, KS, and so on. We’ve added in capabilities in high-frequency analytics like Robust PCA (RPCA), Moving Window PCA, and the capability to detect outliers using Support Vector Data Description (SVDD).

Feel free to come meet with myself and other folks in R&D at SAS Global Forum this year. You will see this exciting new update on full display, and in many whitepapers to follow.

P.S. the complete set of analytics supported in this release are as follows:

Data Wrangling	Modeling
Binning	Logistic Regression
Cardinality	Linear Regression
Imputation	Generalized Linear Models
Transformations	Nonlinear Regression
Transpose	Ordinary Least Squares Regression
SQL	Partial Least Squares Regression
Sampling	Quantile Regression
Variable Selection	Decision Trees
Principal Components Analysis (PCA)	Forest
K-Means Clustering	Gradient Boosting
Moving Window PCA	Neural Network
Robust PCA	Support Vector Machines
	Factorization Machines
	Network / Community Detection
	Text Mining
	Support Vector Data Description

Re: t-test in enterprise miner

jwexler — Thu, 02 Jul 2015 18:12:18 GMT

Hi Wave43. You can certainly use a code node in the context of a larger flow, automatically feeding data into the node for the test. That will work wonderfully. The larger question is what your use case is for running the T Test. If your goal is to generate predictions, then you may be able to run a larger flow using such techniques as regression and neural networks, and generate insights as to how influential your input variable was to your target in the context of a larger flow.

Thanks,

Jonathan

Re: Open Source Integration Node Error

jwexler — Tue, 09 Dec 2014 14:19:37 GMT

Try one last option, make sure when you install the packages you are running as an admin. I had the same problem as you until I did this. if this does not work, try contacting support.sas.com.

Run R as admin and issue:

install.packages("pmml", dependencies=TRUE)

Re: Open Source Integration Node Error

jwexler — Mon, 08 Dec 2014 15:51:13 GMT

There is a usage note at http://support.sas.com/kb/53/794.html

R back-ported support to PMML 4.2 in all archived 3.0.X versions. If you uninstall R and install R 2.15.3 this should solve the problem.

Good luck.

Jonathan

Re: Open Source Integration Node Error

jwexler — Mon, 08 Dec 2014 14:57:43 GMT

Mohamed, can you let us know what version of SAS you are using, along with the version of R installed on your SAS server? There are certain configurations that are not supported, which can cause this error.

Thanks,

Jonathan

Re: Is SAS code available to Bayesian Network?

jwexler — Wed, 20 Aug 2014 15:54:13 GMT

You can run PROC HPBNET which is included with Enterprise Miner. It is a procedure that enables you to generate Bayesian Networks.

Jonathan

Reasons to upgrade to the latest version of SAS Enterprise Miner

jwexler — Wed, 23 Jul 2014 20:06:21 GMT

Good afternoon everyone. Glad to see everyone is participating in this community, the activity has really grown! Based on our recent poll results, I wanted to give you reasons to upgrade to 9.4 Maintenance Release 1 / EM 13.1, which was released in December 2013.

1. 10 new nodes!!! These cover analytic methods including GLM, Support Vector Machines, Clustering, Principal Components, and Time Series (3 new nodes). We also added in the Save Data Node, which enables you to export SAS data sets. Finally, we added in the Open Source Integration node which enables you to embed Open Source R algorithms inside an EM flow.

2. Automatic multi-threading out of the box for many HP nodes, such as Random Forest, Neural Network, SVM, and many others. Take advantage of those multi-core machines!

3. We have documented our High-Performance procedures for use outside of the EM GUI. That means you can write code, taking advantage of the multi-threaded procedures that EM utilizes.

Our next release of Enterprise Miner will be in August, which coincides with 9.4M2 / 13.2. We have taken the time to enhance algorithms such as Random Forest, Neural Networks, Regression, and several others. Look for more announcements.

Best,

Jonathan Wexler

Product Manager - SAS Enterprise Miner

Re: predicted scores in SAS EG

jwexler — Tue, 13 May 2014 14:24:50 GMT

Hi, it appears you asked a related question in a later thread, let us know if there's anything else we can do.

Thanks,

Jonathan

Re: Long Running Time for SAS EM Neural Network

jwexler — Mon, 05 May 2014 14:07:32 GMT

Aha123, have you taken a look at EM 12.3 or even EM 13.1? There is a new High-Performance Neural network node that is automatically multithreaded across the processors on yopur local SAS machine. That should make a difference. Otherwise, let us know what settings you are using.

Thanks,

Jonathan

Re: Variable distribution and diversity measure for selection

jwexler — Fri, 25 Apr 2014 19:25:04 GMT

Wow, great post everyone, very lively today! With respect to any bugs or issues with Enterprise Miner, SAS Tech Support is a great resource for troubleshooting. With respect to coding inside EM, there are many options to customize your flows, including the Code Node, Transformations node, etc... As Reeza stated earlier, some of the finer features may require training and experience. There are many, many macros and macro variables available to add to your coding experience.

Thanks,

Jonathan

Product Manager - SAS Enterprise Miner

Quick SAS Predictive Modeling and Video Tips

jwexler — Wed, 05 Mar 2014 18:00:23 GMT

We have an external site that contains a series of videos that SAS experts have produced that cover our analytic products. For example, if you want to analyze the effect of a Treatment on your outcome, take a peak at the Incremental Response Modeling video.

Video Portal: Statistics and Operations Research

Jonathan

Live SAS Data Mining Q&A March 5 12:30-1:30pm EST

jwexler — Fri, 28 Feb 2014 18:41:45 GMT

Thanks to all of you who participated in the SAS Talks session I hosted last week. It was a blast. To follow up on that event, we will be hosting a LIVE Q&A on the SAS Data Mining Community site on March 5 from 12:30-1:30 PM EST. SAS internal data mining product experts will be on-hand to respond to your questions during the session. We would love to field questions you may have on SAS Enterprise Miner 13.1 or any data mining questions in general.

This will be the first of a series of live Q&A sessions. In the coming weeks, we will be adding interactive chat capabilities to our sessions.

Stop by, and spend your lunch with us!

Thanks,

Jonathan

Re: Enterprise Miner Timeline

jwexler — Fri, 14 Feb 2014 14:41:52 GMT

I have asked one of our SAS project managers to pull together a list. I will post the dates.

This will be a fun test for the community. I will post the dates, and lets see who can produce the chart for everyone! Please share your code.

Jonathan