kipedia.org/wiki/Predictive_Model_Markup_Language">PMML) is an XML format use to share common machine learning model parameters between applications. PMML documents are typically exported from tools such as R after a model has been created..
PMML is very flexible, but sometimes it doesn't declare defaults for missing elements. For example, the "predicted" field is optional (the output field), but Cascading requires it for obvious reasons.
So methods on this class help set sensible defaults if values are missing in the PMML document.
PMMLPlanner in essence maps the PMML elements to Pattern model classes and populates those models with the given parameters. There may not always be a 1 to 1 correlation of elements to Pattern models, for example PMML regression is mapped to Pattern's {@link GeneralRegressionSpec}.
To use, hand PMMLPlanner an PMML XML file, and pass the PMMLPlanner instance to {@link cascading.flow.FlowDef#addAssemblyPlanner(cascading.flow.AssemblyPlanner)}. Use the FlowDef as usual from there. The {@link cascading.flow.FlowConnector#connect(cascading.flow.FlowDef)} will then return a{@link cascading.flow.Flow} that will score your data.
Optionally {@link #resolveAssembly(cascading.pipe.Pipe)} can be used directly, but requires more Cascadingexperience.