Can I run RapidMiner in "batch mode"?

ChrisNelson

I know it can be invoked from a command line and I believe you can pass it a process to load.  What I mean by "batch" is, can I write a script which invokes RapidMiner, specifying a process and a dataset on the command line, and have it export charts for the data without user interaction?

Skirzynski

You can call a script specifiying a process with (resp. for Window your desired script-file is rapidminer.bat in the same directory)

./script/rapidminer //YourRepository/home/test/myProcess"

To specify a dataset you have to use macros. To set a macro from the command line RapidMiner version 5.3 is needed. You can wait for the official release or use the source on sourceforge. Usage:

./script/rapidminer //YourRepository/home/test/myProcess "-Mfile1=/home/to/file1" "-Mfile2=/home/to/file1"

In your process you can use this location in every operator with %{file1} and %{file2}. Unfortunately there is no way to export a chart automatically via the command line - at least as far as i know, but if you are familiar with gnuplot you can use the gnuplot operator.

ChrisNelson

That's disappointing that you can't export charts from the command line.  I realize RapidMiner is a big swiss army chainsaw but all I want it to automate some report/chart generation.  I need access to charting and advanced charting to do that effectively.  If I was going to use gnuplot, I could do that directly, the analysis that I'm doing in RM is quite light weight and I could do it in SQL, perl, or awk.

Skirzynski

I am sorry that we do not meet your requirements. RapidMiner itself concentrates on the design of processes and the analysis of data. For automation and reporting we use RapidAnalytics which can schedule processes automatically and build report-sites from the analysed data (enterprise edition only). Maybe it is worth a try for you.

MariusHelf

If you don't want to invest into the Enterprise Edition of RapidAnalytics, you may want to have a look at the Reporting extension of RapidMiner. Additionally, you could also write the data to disk from within your processes (e.g. with Write CSV) and call gnuplot on that data with the Execute Program operator.

Happpy Mining!
~Marius