guleatoma / pmml

This is modified version of CRAN R package repository. pmml — Generate PMML for Various Models. Homepage: http://zementis.com/

This is a modified version of the pmml package that brings some fixes to the PMML cran package under GPL licence.

This version is still under development. Use at your own risk

To install the package use devtools:

devtools::install_github("guleatoma/pmml")

The package currently brings fixes targeted to the specific use cases listed below.

Modifications brought to the package:

Nested data transformations

In the regular package, having nested data transformations would result in a failure:

Example:

dataBox = WrapData(data)

dataBox <- FunctionXform(dataBox,
                           origFieldName="x1,x2",
                           newFieldName="x12",
                           formulaText="x1 + x2")

dataBox <- FunctionXform(dataBox,
                       origFieldName="x12,x2",
                       newFieldName="x122",
                       formulaText="x12 + x2")                         

would result in a failed transformation to PMML, because the regular package will list in the MiningSchema "x1,x2" and "x2" instead of "x1" and "x2".

Important note on FunctionXform

FunctionXform can bee found in the pmmlTransformations cran package. This function comes with a small error that prevents you from applying FunctionXform several times to dataBox.

So make sure you download the fixed version:

devtools::install_github("guleatoma/pmmlTransformations")

Replacement values for missing values

The regulat package would allow to specify a unique replacement value, this package extends this feature to allow to specify a different replacement value for each variable.

pmml.glm(glm, transforms = dataBox, unknownValue = list("x1" = 0, "x2" = 100))