First Look: IBM SPSS Modeler 16

IBM has recently released IBM SPSS Modeler 16 with three distinct packages:

• Professional with everything for structured data
• Premium adds text analytics, entity analytics and social network analysis
• Gold adds the Decision Management capabilities to the Premium package

IBM SPSS Analytical Decision Management (ADM) will increasingly be included in the IBM SPSS Modeler Gold package rather than sold separately (though it is still available separately – see below for more on new ADM features). I last got updated on IBM SPSS Modeler 15 and IBM SPSS ADM release 7 (see the Product Review here and my thoughts from the launch here).

There are three themes in this release- broader analytics, faster execution and flexible deployment.

Under the heading of broader analytics, IBM SPSS Modeler 16 has added several capabilities:

• A new simulation capability allows you to create simulation data from a small sample of existing data by applying adjustments, such as changed distributions based on how you think the world is going to change. You could also create data to reflect a potential situation to test a corner case or a potential case.
• R support has been added with the addition of R build/score, process and output nodes and through push-back support (where nodes execute in-database) for in-database R engines such as IBM Netezza, Hana and Oracle R. You can also push back to Hadoop using the IBM SPSS Analytical Server product. In addition, IBM added the ability to put R-code behind a custom dialogue to make it easier for non-programmers to integrate R routines into their projects.
• The Premium version’s support for entity analytics has been extended in this release also, adding support for finding related entities, persisting searches and support anonymization of records.
• Plus there are some new chart types and enhanced generalized linear mixed models.

Faster execution is the second area and there’s a couple of changes:

• SQL pushback has been extended as noted to support in-database R and the existing support for in-database mining has been extended to support more of the analytical routines in IBM Netezza and IBM DB2. Support for Hadoop has also been extended in this release.
• Data access and preparation has been improved with more support for aggregate functions, new functionality was added to combine different observations into a single entity, and more support included for spatio-temporal data to track entities in space and time.
• Looping and conditional execution has been added to Modeler and the old proprietary scripting language has been replaced with Python, improving error handling and language support for instance, as well as providing a more familiar language.

Flexible deployment is the third theme:

• Modeler now supports streaming data and is integrated with IBM InfoSphere Streams. The new nodes handle model evaluation, refresh, scoring and updating without suspending data processing.
• There is better integration with IBM Cognos TM1 to support forecasting, where TM1 leverages forecasts built into Modeler, reading and writing directly to the TM1 cube.
• Plus IBM SPSS Analytical Decision Management 8 has been added into the Gold version.

IBM SPSS Analytical Decision Management 8 is now packaged into the Modeler Gold offering and there are a number of new capabilities:

Business scenario analysis has been improved to allow side-by-side comparison of different scenarios, different combinations of business rules, predictive analytics and optimization. Multiple scenarios can be defined and saved and the results- when these scenarios are applied to a sample data set- compared and analyzed. Up to 3 scenarios can be compared at the same time, though any number can be created and saved. Details can be drilled into from the visualization interface.

ADM comes with a number of applications, as well as the ability to create new applications. For ADM 8, IBM has added new applications for Predictive Maintenance and Demand Optimization. Both of these use business rules, predictive analytics and optimization using the integration with CPLEX added in ADM 7.

IBM SPSS Modeler Advantage, the business user interface for Modeler that is included in ADM, has been enhanced in release 8. This has added support for controlling training partitions and algorithm settings (rather than using the defaults). This is designed to keep the option to build a model with a few clicks, while giving some more control over exactly how the model is developed. In addition, a new interactive tree model viewer has been added. All these models can, of course, be modified and extended in Modeler as before.

IBM SPSS Modeler Advantage is also being bundled into different IBM solutions, such as supporting campaign selection using analytics in Unica Campaign.

ADM 8 also adds support for aggregated data across potentially complex hierarchies. The analytical data view allows the developer to link multiple tables and define aggregates or totals for use in potentially multiple Modeler streams or ADM rules. These calculated values can even be calculated using an external ODM ruleset that is then executed as part of ADM or through a remote web services call.

IBM is one of the vendors in our Decision Management Systems Platform Technologies report and you can get more information on IBM SPSS Modeler here.