xgboostRF: xgboost random forests algorithm

In insilico/privateEC: Privacy Preserving Evaporative Cooling Feature Selection and Classification with Relief-F and Random Forests

Description

Scalable and Flexible Gradient Boosting XGBoost is short for “Extreme Gradient Boosting”, where the term “Gradient Boosting” is proposed in the paper Greedy Function Approximation: A Gradient Boosting Machine, by Friedman. XGBoost is based on this original model. This is a function using gradient boosted trees for privacyEC.

Usage

xgboostRF(
  train.ds = NULL,
  holdout.ds = NULL,
  validation.ds = NULL,
  label = "class",
  method.model = "classification",
  cv.folds = NULL,
  num.threads = 2,
  num.rounds = c(1),
  max.depth = c(4),
  shrinkage = c(1),
  objective = "binary:logistic",
  save.file = NULL,
  verbose = FALSE
)

Arguments

train.ds	A data frame with training data and outcome labels
holdout.ds	A data frame with holdout data and outcome labels
validation.ds	A data frame with validation data and outcome labels
label	A character vector of the outcome variable column name
method.model	A character vector of the response variable type for the model
cv.folds	An integer for the number of cross validation folds
num.threads	An integer for OpenMP number of cores
num.rounds	An integer number of xgboost boosting iterations
max.depth	An integer aximum tree depth
shrinkage	A numeric gradient learning rate 0-1
objective	A character vector for the name of the objective function in XGBoost
save.file	A character vector for results filename or NULL to skip
verbose	A flag indicating whether verbose output be sent to stdout

Value

A list containing:

algo.acc

data frame of results, a row for each update

ggplot.data

melted results data frame for plotting

trn.model

xgboost model

elapsed

total elapsed time

See Also

Other classification: epistasisRank(), getImportanceScores(), originalThresholdout(), privateEC(), privateRF(), standardRF()

Examples

num.samples <- 100
num.variables <- 100
pct.signals <- 0.1
label <- "class"
sim.data <- createSimulation(num.variables = num.variables,
                             num.samples = num.samples,
                             pct.signals = pct.signals,
                             label = label,
                             sim.type = "mainEffect",
                             pct.train = 1 / 3,
                             pct.holdout = 1 / 3,
                             pct.validation = 1 / 3,
                             verbose = FALSE)
rra.results <- xgboostRF(train.ds = sim.data$train,
                         holdout.ds = sim.data$holdout,
                         validation.ds = sim.data$validation,
                         label = sim.data$label,
                         num.rounds = c(1),
                         max.depth = c(10),
                         verbose = FALSE)

insilico/privateEC documentation built on May 22, 2020, 5:12 p.m.