By Alex Guazzelli, Ph.D., chief technology officer (CTO), Zementis & Instructor, UC San Diego Extension
The Predictive Model Markup Language (PMML) standard is the lingua franca between data mining applications. With PMML, a model can be created in one system and deployed in another without the need for model recoding. In this case, a model can easily transition from the scientist's desktop where it is built to the operational IT environment where it is put to work.
Traditionally, the task of deploying models took months. That's because a model needed to be recoded every time it moved from one system to another. With the adoption of PMML, companies now take minutes to deploy complex models. This is even more remarkable when the predictive model is composed of a model ensemble, in which several sub-models work together to generate a prediction and models can become quite complex.
In the era of Big Data, the traditional recoding of models is simply a non-starter. To be able to extract value and insight from Big Data, model deployment needs to be as agile as the creation of the data itself, otherwise whenever a model is put to work, it is already stale. With PMML, a complex predictive model containing descriptive, predictive and prescriptive analytics can be created whenever necessary and put to use right-away.
Alex Guazzelli, Ph.D., chief technology officer, Zementis Inc. & instructor, "Predictive Models with PMML", UC San Diego Extension
PMML is supported by most commercial and open-source data mining tools. Companies and tools that support PMML include IBM, SAS, R, Zementis, KNIME, RapidMinder, FICO, StatSoft, Angoss. The standard itself is very mature and its latest release is version 4.2. For more details, see the Data Mining Group website (http://dmg.org) where you can find the language specification.
UC San Diego Extension is also offering an online course about PMML called "Predictive Models with PMML". Using the latest 4.2 version of the software, the course will explore how the PMML language allows for models to be deployed in minutes. You will get to know its business value and the data mining tools and companies supporting PMML. You will also begin to understand the language elements and capabilities and learn how to effectively extract the most out of your PMML code. Alex Guazzelli, Ph.D., the Chief Technology Officer at Zementis, Inc. will be teaching the course.