R/predictor-order.r

In gbm-developers/gbm3: Generalized Boosted Regression Models

# Predictor Order
# 
# Function that calculates the predictor variable order for GBM.
# This is used internally to create the order of variables when growing
# trees in the C++ layer.
# 
# @usage predictor_order(gbm_data_obj, train_params)
# 
# @param gbm_data_obj a \code{GBMData} object.
# 
# @param train_params a \code{GBMTrainParams} object.
#
# @return an updated gbm_data_object that now contains the predictor variable order.
#
# @author James Hickey

predictor_order <- function(gbm_data_obj, train_params) {
  gbm_data_obj$x_order <- apply(gbm_data_obj$x[seq_len(train_params$num_train_rows),,drop=FALSE],2,order,na.last=FALSE)-1
  return(gbm_data_obj)
}