Options

"SVM: How to write weight table"

Legacy UserLegacy User Member Posts: 0 Newbie
edited May 2019 in Help
I run LibSVM and get a tab 'Kernel Model'.

How do I write the content (weight table, support vector table) to a file in a tabular format?
Or alternatively, how would I decode these tables from a model XML file? Is there a documentation for that?

Thanks for any help!                              Stefan
Tagged:

Answers

  • Options
    IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    Hi,

    maybe this is enough:

    just select one cell in the graphical view and press Ctrl-A (select all cells, alternatively you can select them by dragging your mouse), then press Ctrl-C (copy) and insert the contents into Excel / OpenOffice / ...

    From there you can save the file. This of course does not work for automatic processes. For writing those tables into files in a fully automated fashion, you could use the Reporting Engine / Reporting Operators available in the Enterprise Edition only.

    Cheers,
    Ingo
  • Options
    Legacy UserLegacy User Member Posts: 0 Newbie
    Ingo,

    thanks for the quick reply. Indeed, I'm looking for a programmatic solution to allow post-processing. So, I'm probably left with the option of looking into the XML file of the model... (Enterprise seems to create HTML and PDF only? - in any case, I wouldn't have the $$ right now  :()

    Btw., the .mod files lack the closure </object-stream> in RM 1.42

    Kind regards                            Stefan
  • Options
    IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    Hi,

    an additional note: at least the JMySVM operator provides a parameter "calculate_weights" which will deliver the normal vector coefficients. A similar result can be obtained from the SVMWeighting operator. For the support vectors, you could also think of writing a new operator deriving the support vectors from the kernel model. This might be actually easier than extracting the SV from the XML file. Just a few thoughts,

    Cheers,
    Ingo
Sign In or Register to comment.