"how to train a couple of classification models in a loop"

huaiyanggongzihuaiyanggongzi Member Posts: 39 Contributor II
edited June 2019 in Help
At present, I heavily use Rapidminer to train classification model. Generally, the input is a collection of documents having been put into a folder. In other words, for each input, I have to run the workflow once. If I have a couple of input  folders, how can I run this training process in a loop and save the built model with different names. Are there any examples that I can follow? Thanks.

Answers

  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    Hi, are you familiar with macros? In that case, it should be easy to create such a process with the Loop Files operator (enable iterate_over_subdirs parameter, disable iterate_over_files).

    Best regards,
    Marius
  • huaiyanggongzihuaiyanggongzi Member Posts: 39 Contributor II
    Hi, I am not familiar with macros. Are there any tutorials/examples showing the usage of macros? Thanks.
  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    Hi,

    macros are like global variables. You can define them manually with Set Macro. Some operators create macros automatically, amongst them many of the Loop operators.

    Macros can be used in all parameters of all operators by surrounding the macro name with the percent sign and curly braces: %{myMacro}. In that case, before the operator is executed, the %{} construct is replaced with the current value of the macro, then the operator is processed as usual. Just experiment with e.g. Set Macro and Print to Console operators.

    Here are two examples of using macros in a loop. The respective loop operators set some macro values in each iteration of the loop:
    - http://rapid-i.com/rapidforum/index.php/topic,431.msg21409.html#msg21409
    - http://rapid-i.com/rapidforum/index.php/topic,5601.msg19833.html#msg19833

    If you have further questions, just come back here!

    Best regards,
    Marius
Sign In or Register to comment.