The Predictive Model Markup Language (PMML) is an XML-based language aimed at providing a way to exchange different predictive models, for classification or regression purposes, generated using a data mining or machine learning technique. PMML was originally developed by the Data Mining Group (http://www.dmg.org/) in 1997 and its latest version (4.2.1) dates from May 2014.

Even if PMML itself is not a business-oriented language, it is currently possible to generate PMML documents from a variety of well known applications, such as Knime (https://www.knime.org/), or R language (https://www.r-project.org/).

PMML support in Drools is relatively new. It originally started as an experimental module but, with effect from version 6.1, PMML is a first-class citizen of the Drools ecosystem. PMML standard can be used to encode and interchange classification and regression models, such as neural networks, decision trees, scorecards, and others. By adopting PMML, Drools has gained access to a broader...