Skip to content

DecisionTree doesn't handle multiclass input? #134

Closed
Closed
DecisionTree doesn't handle multiclass input?#134
Labels
external pkg issue

Description

@tlienart
tlienart
opened on Nov 18, 2019

Their readme clearly states that it handles cateogircals yet the input_scitype is marked as Continuous and it indeed seems that it does not in fact handle categoricals because it needs to be able to define a isless operation?

Opening an issue on DecisionTree as well for clarifications: https://github.com/bensadeghi/DecisionTree.jl/issues/92

Activity

  1. tlienart
    added
    external pkg issue
    on Nov 18, 2019

  2. ablaom commented on Nov 28, 2019

    @ablaom
    ablaom
    on Nov 28, 2019
    Member

    I think this has now been clarified at the above thread. The DecisionTree input_scitypes could be updated to Table(Continuous, OrderedFactor, Count) but with a comment in the doc-string that coercion of OrderedFactor columns to Count or Continuous may lead to better performance, without affecting outcomes.

  3. ablaom commented on Nov 28, 2019

    @ablaom
    ablaom
    on Nov 28, 2019
    Member

    To be clear, Multiclass is not strictly supported. If such data is coerced to one of the other ordered types, by choosing an ordering, then the outcome of training will depend on that choice.

  4. ablaom
    closed this as completedon Nov 28, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    external pkg issue

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions