45 lines (33 loc) · 2.69 KB

treeshap (development version)

treeshap 0.4.0

Fixed xgboost api and add depends R>=4.1.0 as in #51 (#51). Thanks @kapsner
Fix ranger.unify for probability forests as in #43 (#43). Thanks @ck37
Added support for GPBoost as in #41 (#41). Thanks @fabsig

treeshap 0.3.1

Fixed code examples in lightgbm.unify.

treeshap 0.3.0

Fixed ranger_surv.unify operation for predictions in form of survival and cumulative hazard functions.
Added model_unified_multioutput and treeshap_multioutput classes for multi-output models and their explanations.
Improved documentation of ranger_surv.unify.

treeshap 0.2.5

Removed catboost.unify function (as the catboost package is not on CRAN); it is available on a separate branch
Fixed randomForest.unify for classifiers (#12, #23)
Implemented consolidated (generic) unify function (#18)
An error is thrown when the data passed to the unify or treeshap functions contain variables that are not used by the model (#14)
Added implementation for random survival forests created using ranger (#22, #26)
Fixed GitHub Actions, check and test issues (#25, #29)
Fixed issues with documentation and examples
Changed use of bitwise '|' to logical '||' with boolean operands in C++ files

treeshap 0.1.1

Fixed plot_contribution when max_vars is larger than the number of variables (#16)

treeshap 0.1.0

Rebuilded treeshap function so it now stores observations and whole dataset
Rebuilded all unifiers so they require passing data.

treeshap 0.0.1

Made package pass all checks
Fixed infinite recursion issue in ranger (see commit)
If there is no missing value in the model, unifiers return NA for Missing column (see commit)

treeshap 0.0.0.9000

treeshap is now public
Implemented fast computations of tree ensemble shap values in C++
Implemented unifiers for catboost, lightgbm, xgboost, gbm, ranger and randomForest