PMML Compiling Predictor

This node translates a PMML model into Java scoring code and compiles it into bytecode. It then executes the code with the data from the second port. The precompilation can significantly speed up the scoring process for huge amounts of data. For small data sets the translation process might be too big of an overhead. Currently supported PMML models are all those generated by KNIME plus Naive Bayes:

• Regression
• SVM
• MLP
• K-Means Clustering
• Decision Trees
• Naive Bayes
• Rulesets
• Ensembles of the above

Options

Change prediction column name

When set, you can change the name of the prediction column.

Prediction Column

The possibly overridden column name for the predicted column. (The default is: Prediction (trainingColumn).)

Append columns with normalized class distribution

Shows the normalized class distribution for each prediction (If applicable, eg for classification models).

Suffix for probability columns

Suffix for the normalized distribution columns. Their names are like: P (trainingColumn=value).

Views

This node has no views

Workflows

• k_010_apply_to_new_dataKNIME Hub
• kn_example_decision_treeKNIME Hub
• M 010 - Decision Tree and all the ExportsKNIME Hub

Links

• Export PMML that outputs class probabilitiesby mlauber71 on 2019-01-22

  forum.knime.com/p/53278

Developers