R Predictor

Allows execution of an R script from within KNIME. The resulting table should contain predictions about data attributes based on a model generated with an R Learner node.

More details about R:
http://www.r-project.org/
Short-refcard.pdf

Options

R Command

R Snippet
The code being executed in R. New input data is available as a data frame named knime.in, while the desired model must be available as a variable in the workspace from a previously executed R Learner node.
Flow Variable List
The list contains the flow variables that are currently available at the node input. Double clicking any of the entries will insert the variable at the current cursor position.
Column List
The list contains the column references that are currently available at the node input. Double clicking any of the entries will insert the respective identifier at the current cursor position.

Templates

Category
A template category defined by the author.
Title
Names of the available templates.
Description
A brief overview of the purpose of the template.
Preview
The code to be applied to your expression editor.

Advanced

NaNs
Convert NaN and +/-Infinity output to missing cells. Enable this for backwards compatibility with the behavior before version 2.10.
Send Row Names to R
Whether to send the row names from KNIME to R. Disabling this can improve performance with very large tables. Row names of knime.in will instead be default R row names: 1:n.
R type for knime.in
By default knime.in is data.frame. For large input tables, using data.table is likely more memory efficient especially while sending the input data to R and may therefore be faster. Please be aware that the support for data.table is experimental (!) and requires the "data.table" R package to be installed. It may not work for some R scripts, may slightly change or even be removed entirely in a future version.
Send batch size
The input of the R node is sent in row batches (default, 10 000 rows at a time). This is the amount of rows KNIME will retain in memory at any given time. For large tables, a higher batch size may be faster, but will require additional memory. Be aware, that R will likely keep all of the input table in memory and this only helps with memory during transmission of data to R.
Overwrite default path to R home
Check this option to select a path to the home of a R installation that is different from the path configured on the R preference page. The path must point to the root folder of the R installation tree. You can either specify the path directly or use the R installation in a conda environment. The conda environment can be specified using a conda environment flow variable from e.g. a “Conda Environment Propagation” node.

Input Ports

Icon
R Model Input
Icon
Data Input

Output Ports

Icon
Predicted data output

Views

R Std Output
The text sent to standard out during the execution of R executable as well as the text of the .Rout file generated by R.
R Error Output
The text sent to standard error during the execution of R executable as well as the text of the .Rout file generated by R. (If it appears in gray, it's the output of a previously failing run which is preserved for your trouble shooting.)

Workflows

Links

Developers

You want to see the source code for this node? Click the following button and we’ll use our super-powers to find it for you.