xg_gs: Grid search

In ArnaudBu/ezXg: Easy xbgoost wrapper

Description

xg_gs use coordinate descent (https://en.wikipedia.org/wiki/Coordinate_descent) in order to select the best set of parameters for an xgboost model. At the end of the coordinate descent algorithm, a full search on each of the individual parameter vectors is made in order to potentially improve the selection.

Usage

xg_gs(data, eta = c(0.05, 0.1, 0.15, 0.2, 0.25, 0.3), gamma = c(0, 0.1,
  0.2, 0.3, 0.4, 0.5), max_depth = c(1, 3, 4, 5, 6, 8, 10, 12, 15),
  colsample_bytree = c(0.3, 0.4, 0.5, 0.7, 0.8, 0.9, 1),
  min_child_weight = c(1, 3, 5, 7), nrounds = 100, nthread = 2,
  cv = 5, seed = 1, verbose = TRUE, objective = "auto")

Arguments

data	Object. A data structure created by the call of the xg_load_data function.
eta	Numeric Vectors. Eta parameter list for grid search. See xgb.train for more details.
gamma	Numeric Vector. Gamma parameter list for grid search. See xgb.train for more details.
max_depth	Numeric Vector. Max_depth parameter list for grid search. See xgb.train for more details.
colsample_bytree	Numeric Vector. Colsample_bytree parameter list for grid search. See xgb.train for more details.
min_child_weight	Numeric Vector. Min_child_weight parameter list for grid search. See xgb.train for more details.
nrounds	Numeric. Nrounds parameter for xgboost calibration. See xgb.train for more details.
nthread	Numeric. Nthread parameter for xgboost calibration. See xgb.train for more details.
cv	Numeric. Number of folds in cross validation. Needs to be more than 2.
seed	Numeric. Seed for computation reproducability.
verbose	Logical. Verbose parameter for grid search.
objective	Character. Objective function for the optimization. . Eta parameter for xgboost calibration. See xgb.train for more details. Can be set to auto in order to let the function choose the better model regarding the output variable.

Value

The optimization results with the following fields:

• param: the optimal set of parameters.

• err: the error associated to the optimal parameter set.

• results: the history of the results for the cross-validation with all the tested sets of parameters.

Examples

d <- xg_load_data(system.file("extdata", "titanic.csv", package = "ezXg"),
               inputs = c("Pclass", "Sex", "Age", "SibSp",
                          "Parch", "Fare", "Embarked"),
               output = "Survived",
               train.size = 0.8)
t <- xg_gs(d)

ArnaudBu/ezXg documentation built on Oct. 30, 2019, 4:59 a.m.