The Statistica Rapid Deployment of Predictive Models module efficiently generates predictions from previously trained models using industry-standard PMML deployment code. It can process single or multiple PMML files to compute predicted values or classes for test data based on trained models.

PMML is an XML-based language for encoding information from data mining projects. This module is especially useful for generating predictions for a large number of observations as it passes through the data once, storing only the data for a single observation at a time.

This article shows the process of Evaluating / Scoring of the latest PMML 4 models via Rapid Deployment Node in Statistica Application.

Windows

The Rapid Deployment node has a new checkbox “Evaluate latest version PMML” that is available to evaluate the PMML 4 generated from the modeling nodes.

Currently, some Statistica modeling nodes generate two PMML versions.
The checkbox option selects between PMML 4 (Standard/Compliant PMML) and the PMML (Legacy Statistica PMML) versions and loads the appropriate model based on the checkbox selection on Rapid Deployment. 
The checkbox is deselected by default for backward compatibility reasons and for supporting the evaluation of the legacy Statistica PMML model with this Statistica release.

A new checkbox option called,  “Evaluate the latest version PMML” was included on the Quick tab of Rapid Deployment node.
On executing the Rapid Deployment node with this option selected, the system will invoke the jpmml-evaluator.jar library and evaluate the PMML model on the PMML 4 tab of the upstream PMML nodes, and generate summary/downstream outputs.

Checkbox available for evaluation PMML 4 :



Autodetection of PMML 4 model over the legacy ones :




The example consisting of screenshots below shows the Rapid Deployment node with summary predictions and other results for the latest PMML version of the Boosted Classification Trees model.

a. Rapid Deployment by selecting "Evaluate the latest version PMML" on CrediScoring Data
 

b. Reports / Model evaluation done on PMML 4 


We have added the workspace demonstrating the PMML 4 evaluation using Boosted Classification Trees.

https://support.tibco.com/s/article/PMML-versions-supported-by-Rapid-Deployment-Node-in-Statistica
https://support.tibco.com/s/article/How-to-deploy-a-PMML-script-of-a-data-miner-model-through-Rapid-Deployment-node-in-workspace