Clementine 12.0, the latest version of the vendor's predictive analytics software, extends automated modeling capabilities. In Clementine 11.1, developers could build and evaluate binary (yes/no) models in one click. Clementine 12.0 adds this capability for models that yield continuous variables. In addition, Clementine 12.0 has an "Ensemble" modeling node that lets users blend models to achieve more accurate predictions.
"Certain types of models tend to have certain biases," says Richard Hren, SPSS director of product marketing. "Blended modeling is a hot topic because you can get a better model by combining several models into one. The Ensemble modeling node will combine models and average out the differences automatically."
The Clementine 12.0 release also incorporates improved data visualization capabilities via a Graph Board that lets users develop bar charts, scatter plots, concept maps and linkage maps. These visualizations help make predictive analyses readily understandable to business users. In addition, built-in interaction features let users highlight, rope or otherwise select data for deeper analysis and visualization.
Text Mining for Clementine 12.0 is designed to extract concepts, sentiments and relationships from textual data such as e-mail messages, blogs, RSS feeds or surveys. Putting that insight to use, the software is designed to work with SPSS's Dimensions survey software, and it is integrated with SPSS Predictive Enterprise Services, so text mining results can be combined with the predictive power of Clementine for broader and more accurate insight.
"As we've made text mining easier to use and more tightly integrated within our predictive analytics portfolio, we've seen user adoption and acceptance flourishing," says Hren. "Lots of vendors are doing text mining and content classification, but for us, the real power is in linking that to the analytical space."
As an example, SPSS points to European cable provider Cablecom, which uses Text Mining for Clementine to analyze call center notes fields. The company then combines those results into its predictive analyses for better predictions and more effective efforts to mitigate churn.
Among the other refinements to Clementine 12.0 and Text Mining for Clementine, the software is said to take better advantage of multithreading, multicore processors and load balancing in clustered server environments. As a result, processing speeds in clustered deployments are reportedly as much as 10 to 20 times faster than in older versions of the software.