Intro to Machine Learning Q&A, Slides, and On-Demand Recording

[#1]

Watch this Ask the Expert session to learn what exactly machine learning is and what SAS offers for machine learning.  

Watch the webinar

Join Melodie Rush as she discusses what machine learning is and shows you examples of specific machine learning techniques, including random forest, gradient boosting, support vector machine, neural networks and K-means. Watch this webinar to:

• Learn terminology for machine learning.
• Get an introduction to machine learning algorithms.
• See a software demo on both SAS^® Enterprise Miner^™ and SAS^® Viya^®.

The questions from the Q&A segment held at the end of the webinar are listed below and the slides from the webinar are attached.

What is output?

Output is the results from running an algorithm.

Which of these Machine Learning techniques are available in the SAS/STAT package?  Where can I find info on syntax for running these routines?

Both Regression and Decisions Trees are available in SAS/Stat. The documentation for these procedures can be found here https://support.sas.com/rnd/app/stat/procedures/Procedures.html

In ensemble modelling, what type of sub-models are used?

Depending on what tool you are using, will determine the models you can use. SAS Enterprise Miner and SAS Visual Data Mining and Machine Learning allows you to ensemble any of the modeling algorithms included in the product.

For model comparison, did you import just the results of R decision tree? Did you need to convert the R-results into a table before importing into SAS?

Both SAS Enterprise Miner and SAS Visual Data Mining and Machine Learning use the integration node to pass data and instructions back and forth between R and themselves. For the model comparison the predictive probabilities from the models are used to create the model comparisons.

In the unsupervised learning, explain what choices the user is actually determining or is the computer doing it all?

For unsupervised learning, the user would be able to choose the input for the algorithm and any parameters the algorithm has available.

Does SAS EG support code for supervised machine learning? Can PROC Forest be used in SAS EG?

SAS Enterprise Guide can run SAS Procedural code for any licensed product you have.

Is AICC available as a criterion in model comparison?

AICC is available in SAS Enterprise Miner for model comparison.

How are data imputed as you showed earlier? Is it using multiple imputation techniques?

In my examples I used simple imputation techniques; count for categorical columns and mean for continuous. There are many other techniques available. You can review these techniques for SAS Enterprise Miner HP Impute Node here and for the Impute Node here 

For SAS Visual Data Mining and Machine Learning Pipelines Imputation node Chapter 14 covers the imputation methods. 

https://go.documentation.sas.com/api/docsets/vdmmlref/8.5/content/vdmmlref.pdf

Difference between data mining and machine learning?

In the presentation this Venn diagram was shared. Machine Learning and Data Mining overlap with many of the algorithms used. Machine Learning will typically be part of a larger AI project that is tasked with automating a process.

 

Other than PROC SURVEYREG and SURVEYLOGISTIC are there ML methods for weighted Survey data?

The complete list of Survey procedures can be found here https://support.sas.com/rnd/app/stat/procedures/SurveyAnalysis.html I would put SURVEYIMPUTE and SURVEYPHREG as possible inclusions to your list for ML methods.

How does SAS' ML models compare against models in R or Python?  I noticed you had an "R - Decision Tree" in Enterprise Miner and it was pretty good against the other models.

Every set of analytical data will have a best algorithm. The data scientist, data analyst, or statistician will use their skills to decide what is best for their data. The data I used for the presentation does well with trees and a well-tuned Gradient Boosting model is really hard to beat based on misclassification as the criterion. If I changed my criterion it is very possible that another modeling technique would work better. Or if I do some clever feature creations perhaps a different algorithm will be better. At the end of the day, you need to know what your outcome is. Is it to have a pretty good algorithm or the best algorithm? And what does the best mean? Good fit, accuracy, precision, or something else? For most projects it boils down to money. Does the model make money and/or save money? And what are the costs of the model to maintain and deploy.

What is the best machine learning algorithm?

The best machine learning algorithm is the one that works best with your data for you to achieve your outcome.

What SAS products allow me to do clustering?

SAS/Stat, SAS Enterprise Miner and SAS Visual Statistics

Do I have to have SAS Visual Data Mining and Machine Learning product to do Machine Learning?

It will depend on what algorithm you want to use. SAS/STAT, SAS Enterprise Miner, SAS Visual Statistics and SAS Visual Data Mining and Machine Learning all have Machine Learning algorithms included.

Recommended Resources

Machine Learning – What it is and why it matters

SAS Visual Data Mining and Machine Learning on SAS Viya – free trial

SAS Enterprise Miner Learn & Support

Which machine learning algorithm should I use?

Want more tips? Be sure to subscribe to the Ask the Expert board to receive follow up Q&A, slides and recordings from other SAS Ask the Expert webinars.