v0.4.3

MLJDecisionTreeInterface v0.4.3

Diff since v0.4.2

• Bump compatibility requirement, CategoricalArrays="1"

Merged pull requests:

• Bump CategoricalArrays compat to 1.0 (#60) (@ablaom)
• For a 0.4.3 release (#61) (@ablaom)

v0.4.2

MLJDecisionTreeInterface v0.4.2

Diff since v0.4.1

Merged pull requests:

• Improve handling of row tables by reformat (#57) (@ablaom)
• For a 0.4.2 release (#58) (@ablaom)

v0.4.1

MLJDecisionTreeInterface v0.4.1

Diff since v0.4.0

(enhancement) Include feature names when printing a tree (#54)

Merged pull requests:

• Fix some doc strings, add StatisticalMeasures.jl to test deps (#53) (@ablaom)
• Print the feature names in report.print_tree() (#54) (@adarshpalaskar1)
• For a 0.4.1 release (#55) (@ablaom)

Closed issues:

• Tag a new breaking release and yank 0.3.1 (#48)

v0.4.0

MLJDecisionTreeInterface v0.4.0

Diff since v0.3.0

Closed issues:

• Increase default number of trees in forests from 10 to 100? (#43)
• Wrap trees using DecisionTree.wrap for more convenient display (#45)
• Version 0.3.1 broke tests in MCMCChains (#47)

Merged pull requests:

• Change n_trees default to 100 (#49) (@ablaom)
• Wrap trees for better display (#50) (@ablaom)

What's Changed

• Change n_trees default to 100 by @ablaom in #49
• Wrap trees for better display by @ablaom in #50

Full Changelog: v0.3.1...v0.4.0

v0.3.0

MLJDecisionTreeInterface v0.3.0

Diff since v0.2.5

• (breaking) Require DecisionTree.jl 0.12.0. This means seeding of individual RNGs in a random forest has changes, which in turn changes the outcomes of training for both random forest models (JuliaAI/DecisionTree.jl#198)
• (breaking) Reverse the role of key and value in the encoding dictionary returned as part of fitted parameters in a DecisionTreeClassifier (#37)

Closed issues:

• Reverse roles of key and value in encoding returned in fitted params (#19)
• DecisionTreeRegressor and DecisionTreeClassifier do not allow n_subfeatures=-1. (#33)

Merged pull requests:

• update docs per #33 (#34) (@john-waczak)
• Bump DecisionTree.jl to 0.12 (#35) (@rikhuijzer)
• For a 0.3.0 release (#36) (@ablaom)
• Reverse roles keys <--> values in encoding for classifier (#37) (@ablaom)
• For a 0.3.0 release - Take 2 (#38) (@ablaom)

v0.2.5

MLJDecisionTreeInterface v0.2.5

Diff since v0.2.4

Merged pull requests:

• MLJ model interface compat fix (#31) (@ablaom)
• For a 0.2.5 release (#32) (@ablaom)

v0.2.4

MLJDecisionTreeInterface v0.2.4

Diff since v0.2.3

(new feature) Implement feature_importance accessor function for all models, with type of importance controlled by new hyperparameter feature_importance. This can be :split or :impurity (default). See the docstring for RandomForestClassifier for an example. (#28) @john-waczak @OkonSamuel

Closed issues:

• Incorporate feature importance (#24)

Merged pull requests:

• Adding feature_importances to report for DecisionTree models (#28) (@john-waczak)
• For a 0.2.4 release (#29) (@ablaom)
• Update docstring (#30) (@ablaom)

v0.2.3

MLJDecisionTreeInterface v0.2.3

Diff since v0.2.2

Closed issues:

• MethodError: no method matching metadata_model (#20)

Merged pull requests:

• Fix build badge on readme. (#22) (@ablaom)
• Implement StableRNGs throughout tests to fix reproducibility problems in CI (#26) (@ablaom)
• For a 0.2.3 release (#27) (@ablaom)

v0.2.2

MLJDecisionTreeInterface v0.2.2

Diff since v0.2.1

Merged pull requests:

• For a 0.2.2 release (#21) (@ablaom)

v0.2.1

MLJDecisionTreeInterface v0.2.1

Diff since v0.2.0

Closed issues:

• Remove (non standard) smoothing of probabilistic predictions (#11)

Merged pull requests:

• For a 0.2.1 release (#18) (@ablaom)