Requests for waffles_transform

Help
Anonymous
2012-06-22
2012-09-14

  • Anonymous
    2012-06-22

    A couple requests--not a big deal as it's pretty easy to code these myself
    outside of waffles, but would help with completeness:

    1. It would be handy to have an option in waffles_transform along the lines of dropmissingvalues that would drop any column whose values are all identical.
    2. Rebuild the list of class values for a nominal attribute from the values in that column.

    On both of the above, my issue is that I'm finding that after I do a split on
    an arff dataset, I have some redundant columns/attributes.

    Thanks!

     
  • Mike Gashler
    Mike Gashler
    2012-06-22

    1 sounds pretty easy to implement. I don't quite understand #2. Could you

    elaborate on what it does and how that differs from what you would prefer?

     

  • Anonymous
    2012-06-22

    I'd rephrase #2 as: Remove class values for categorical attributes that are no
    longer present in that column.

    For instance, suppose you have file1.arff that starts out with the attribute:

    @ATTRIBUTE people {Alice, Bob, Charlie}

    Now I do some sort of split on the file1.arff producing file2.arff. In
    file2.arff, I only have "Bob" and "Charlie" in this column. What I'd like is a
    command in waffles_transform that will reset the ATTRIBUTE list to just say

    @ATTRIBUTE people {Bob, Charlie}

    Thanks,

    SeattleDad

     
  • ok, I added both features. Here is example usage:

    waffles_transform drophomogcols mydata.arff

    and

    waffles_transform dropunusedvalues mydata.arff

     


Anonymous


Cancel   Add attachments