What you should know about SAS Factory Miner


On July 14, 2015 SAS released SAS Factory Miner, our latest advanced data mining and machine learning product.  This new product provides automated predictive modeling by segment in a high velocity grid enabled environment, allowing modelers to run hundreds of models within minutes and find champion models by segment quickly.  Below is a short description of how it works and a top ten list of things to know about SAS Factory Miner.

Data sources for modeling are defined in metadata, just like when using SAS Enterprise Miner.  Target variables and segment variables are identified, variable roles and levels are defined and once the data source is saved in metadata, it is available to everyone in your organization.

You start a new project by providing a name, location and selecting a data source.

Factory Miner: New-Project

The variables in the data will display in a table and you can choose to change the data properties of the variables if you would like to. You can also select a variable and view a distribution of the variable.


The next step is to build a segmentation profile (click on Build Profile in upper right) which partitions the data into training and validation data sets and applies a segmentation scheme.  You can also filter segments to be included or excluded from modeling.


After a segmentation profile is created, you can choose the models you want to run from the available templates, or build your own templates. The templates provided are:

  1. Bayesian Network Model
  2. Decision Tree Model
  3. Generalized Linear Model
  4. Gradient Boosting Model
  5. Neural Network Model
  6. Random Forest Model
  7. Regression Model
  8. Support Vector Machine Model


Think of these templates as the equivalent of SAS Enterprise Miner process flows. Each process flow is made up of components which are the equivalent of nodes in EM, such as Impute, Transform variables or Variable selection.  You select the components you want to be part of the process flow and you can edit the properties of each component.  Once edited, you can save the template as a global template that can be shared by everyone in your organization.


Once the model templates are selected and any changes made to the components, select the Run button.  The segmented models run asynchronously across a grid. The results window displays and the left side (below the plots) shows a dynamically changing status as the model runs are completed.  A variable importance plot is displayed in the upper left and a model performance plot based on a chosen statistic is shown in the upper right.


At this point, you can drill into the results for a segment, view model graphs for the champion model and choose to make changes to the segmented models and rerun.


The last step is to put the champion models into production with the aid of SAS Model Manager.  That happens by first registering the champion models to metadata.  On the Projects screen, select the project and click on the Register Project icon on the toolbar. You can register all models or only the champion models.


The new automated predictive analytics model factory found in SAS Factory Miner provides benefits that include:

  1. Flexible environment
  2. Reusable components
  3. Easy to collaborate with others in an organization.

Below is a top ten list of things you should know about SAS Factory Miner……enjoy!

  1. It is a browser-based HTML5 interface
  2. It is an “Analytics Factory” providing big data automated modeling by segment
  3. It uses HP procedures and Grid to satisfy needs for a higher velocity model development environment
  4. Give your new project a name, select a data source, assign your target and segment variables and you are on your way to automated model building
  5. Data sources are registered in metadata once and reusable by all
  6. Model templates are provided, which are customizable and can be saved and shared with others
  7. Modeling results allow you to look at results across segments as well as compare models within segments
  8. SAS Enterprise Miner is required
  9. It is integrated with SAS Enterprise Decision Management (and therefore SAS Model Manager)
  10. It really is a data scientist in a white box. Run on autopilot when appropriate and take control when needed….

Want to learn more about how you can build better models with SAS Factory Miner?

Get more information and view a demo here.



About Author

Marjorie Shelley

Principal Technical Training Consultant, SAS Professional Services

Marjorie Shelley is a Principal Technical Training Consultant in Global Enablement and Learning, where she focuses on creating training assets that cover all areas of SAS analytics. She has been with SAS for over 20 years, the first 10 spent as a Systems Engineer in Sales and the last 10 in internal training.

Leave A Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Back to Top