Taking SPSS Modeler to the cloud and beyond at Insight 2015
The SPSS Modeler crowd might not be the largest contingent at IBM Insight, but Insight 2015 kept us busy even so, offering a related presentation in every time slot. I’ve been using Modeler for more than 15 years, but Insight reminded me that I still haven’t finished learning about it. Here are a few Modeler-related topics that loomed large on my radar during Insight this year.
Tackle big data at scale
The newest release of SPSS Modeler, version 17.1, is more than a mere patch. In particular, many of its features focus on big data, which—as we learned during numerous presentations at Insight 2015—is set to be the next major theme for Modeler.
A noteworthy feature of Modeler 17.1 is the addition of algorithms that scale to big data when used in combination with SPSS Analytic Server, among them a new version of Random Trees. Unlike implementations you may have seen done in R, this implementation is not only native, but also scalable. For more, watch the Big Data & Analytics Hub for a post dedicated to SPSS Modeler, big data and Spark.
Armand Ruiz and his colleagues have been trying to spread the word about IBM Bluemix and how it allows the deployment of Modeler streams on the cloud. Of course, as with any new technique, the best way to master it is to practice it, and users of Modeler can do just that thank to a new SPSS Modeler and Predictive Analytics fundamentals course made available through the Big Data University website. Taught by Mikhail Lakirovich, Greg Filla and Armand Ruiz, it concludes with an example of a Bluemix deployment and helps start Modeler users on the path to leveraging cloud capabilities.
Make pasting the best choice on the menu
Jos den Ronden and Jim Mott, both of the SPSS curriculum team, brought a combined 74 years of SPSS experience to their workshop at Insight, in which they demonstrated how to use the menus in Modeler to prototype a Python script. As attendees learned, Modeler users can produce a complete script without coding, though within certain limitations.
When creating a script in Modeler, basic if/then logic and simple loops are available using the menus in version 16 or later. (Try it for yourself by right-clicking a node in Modeler: You’ll find looping and conditionals in the context menu.) However, for more ambitious scripts, Modeler users can paste in a script in much the same way that users of SPSS Statistics can, then use their Python skills to add the features they desire.
Small wonder, then, that Steve Poulin of Aviana Global drew a crowd at 8:30 AM to watch him use Python to program a complex loop in SPSS Modeler. By doing so, he was able to push more than 30,000 forecasts out to a table from which his colleague retrieved them, pulling them into Cognos TM1 for Pabst.
Let extensions do the heavy lifting
R and Modeler got their own Insight workshop thanks to Armand Ruiz and Richard Cohen, who demonstrated how to combine R code, the Custom Dialog feature and R nodes to produce a complete solution—in the case of their example, a compelling demonstration of geocoding. As attendees learned, doing so is surprisingly easy. Better still, by using any of more than 100 premade solutions, Modeler users can avoid doing extra work and get right to business—all thanks to the IBM SPSS Predictive Analytics Gallery.
To discover what the latest version of SPSS Modeler has to offer your organization, learn more predictive analytics, statistical analysis and IBM analytics resources.