Fasten your seat belts - it's time to put the pedal to the metal. In this segment, we're going to burn rubber and unleash the full fury of the advanced RPM model on the unsuspecting burglary data!
We will pop the hood and take a sneak peek at the drive train.
If any State Troopers have their radar guns pointed here, please be aware there are no limits enforced on our SAS test track. Well okay, maybe some confidence and prediction limits.
To really get an idea of the handling and responsiveness of Enterprise Miner and RPM, why not sign up for a test drive in one of our high-performance classes where you can be at the controls. You will not be disappointed.
Here's Part 1 and Part 2 of my video and blog on descriptive modeling using SAS. Stay tuned for Part 4!
You can also learn more in our courses, Rapid Predictive Modeling for Business Analysts and Applied Analytics Using SAS Enterprise Miner.
2 Comments
Thank you for this fantastic series of posts. In the third video, you mentioned you could run new data through your model if you only have demographic data. I would like to see this in action using RPM. i.e. How does one package the model to use for new data?
Also, any discussion on validating models over time is also appreciated. I'm thinking of relationships in the data that may have changed suddenly (or gradually) over time.
Jared, thanks for your questions! After running the RPM task once, look for “Registration” in the menu where you set your model properties. Fill in the fields to register the model in the SAS Metadata Repository.
To run new data through your model, just open the new data in Enterprise Guide, and then select Tasks-Data Mining-Model Scoring to launch the scoring task. This will let you select your RPM model, and your new data will be scored right there.
As far as validating models over time, SAS Model Manager is designed to do just that. It tracks changes in the distribution of your input variables as well as your target over time. It will show you any variables that deviate from their original distribution, and by how much. Of course there are lots more details on Model Manager, but this is how people track models over time.