ODRF (development version)
- Added linear model tree. Specifically, use parameter Xsplit as the splitting variable for ODT and fit a linear model for each split using the function “gmlnet”. The corresponding parameter is split=“linear”.
- Added the ensemble of ODT-based boosting trees，denoted by ODBT.
- Changed the parameter “leafnode” to “type” in the function predict.ODT(), and for classification tasks, added category probability output.
- Optimized some other known issues.
ODRF 0.0.4
CRAN release: 2023-05-28
- Fixed function VarImp(), adding the method of measuring the importance of variables with node purity, and now VarImp() can be used for both class ODT and ODRF.
- When the argument “Xcat ! = 0”, i.e., the category variable in predictor X is transformed to one-of-K encode. however for the argument “NodeRotateFun=‘RotMatRF’ (‘RotMatRand’)“ run error, we have now fixed it.
- Added predicted values of training data for class ODT and ODRF.
- Fixed issue related to function predict.ODRF() when argument “weight.tree = TRUE”.
- Optimized some other known issues.
ODRF 0.0.3
CRAN release: 2023-03-16
- The function predicate.ODT() runs error when ODT is not split (depth=1), and we have fixed this bug.
- We have fixed the function predict.ODRF() with arguments numOOB and weight.tree related issues.
- We have fixed the functions plot.ODT(), VarImp() and plot.VarImp().
- We have fixed the argument ‘lambda’ of the functions ODT() and ODRF().
ODRF 0.0.2
CRAN release: 2023-02-28
- We have now explained CART and Random Forest in the description text.
- We have changed the Date field to a more recent date.
- We have now exported the functions RandRot() and defaults(), and no longer need ODRF:::
- We have removed par from plot.VarImp() and added on.exit to plot.prune.ODT(), and checked the code to make sure that it does not change the user’s options, including par or working directory.
- We have removed the random seed number in functions ODRF(), poune.ODRF(), online.ODRF() and plot_ODT_depth().