varimp_stratified_rf: Heuristic on variable importance
In david-cortes/StratifiedRF: Builds Trees by Sampling Variables in Groups

Description Usage Arguments Details Value Examples

Heuristic on variable importance, taken as averages from the variable importances calculated for each tree.

1	varimp_stratified_rf(model, metric = "usage", agg_type = "simple")

`model`	A stratified_rf model.
`metric`	How to calculate the variable importance from each tree. Either "usage" or "splits".
`agg_type`	How to aggregate the variable importances obtained from each tree. Either "simple" for a simple average, or "weighted" for an average weighted by each tree's accuracy.

Methods are taken directly from the C5.0 trees. Currently doesn't support permutation tests.

A named data frame with the importance score of each variable, sorted from largest to smallest.

data(iris)
groups <- list(c("Sepal.Length","Sepal.Width"),c("Petal.Length","Petal.Width"))
mtry <- c(1,1)
m <- stratified_rf(iris,"Species",groups,mtry,ntrees=2,multicore=FALSE)
varimp_stratified_rf(m)