Bayesian Optimization for XGboost

This function estimates parameters for xgboost based on bayesian optimization.

xgb_opt(train_data, train_label, test_data, test_label, objectfun, evalmetric,
  eta_range = c(0.1, 1L), max_depth_range = c(4L, 6L),
  nrounds_range = c(70, 160L), subsample_range = c(0.1, 1L),
  bytree_range = c(0.4, 1L), init_points = 4, n_iter = 10, acq = "ei",
  kappa = 2.576, eps = 0, optkernel = list(type = "exponential", power =
  2), classes = NULL)

Arguments

train_data	A data frame for training of xgboost
train_label	The column of class to classify in the training data
test_data	A data frame for training of xgboost
test_label	The column of class to classify in the test data
objectfun	Specify the learning task and the corresponding learning objective reg:linear linear regression (Default). reg:logistic logistic regression. binary:logistic logistic regression for binary classification. Output probability. binary:logitraw logistic regression for binary classification, output score before logistic transformation. multi:softmax set xgboost to do multiclass classification using the softmax objective. Class is represented by a number and should be from 0 to num_class - 1. multi:softprob same as softmax, but prediction outputs a vector of ndata * nclass elements, which can be further reshaped to ndata, nclass matrix. The result contains predicted probabilities of each data point belonging to each class. rank:pairwise set xgboost to do ranking task by minimizing the pairwise loss.
evalmetric	evaluation metrics for validation data. Users can pass a self-defined function to it. Default: metric will be assigned according to objective(rmse for regression, and error for classification, mean average precision for ranking). error binary classification error rate rmse Rooted mean square error logloss negative log-likelihood function auc Area under curve merror Exact matching error, used to evaluate multi-class classification
eta_range	The range of eta
max_depth_range	The range of max_depth
nrounds_range	The range of nrounds
subsample_range	The range of subsample rate
bytree_range	The range of colsample_bytree rate
init_points	Number of randomly chosen points to sample the target function before Bayesian Optimization fitting the Gaussian Process.
n_iter	Total number of times the Bayesian Optimization is to repeated.
acq	Acquisition function type to be used. Can be "ucb", "ei" or "poi". ucb GP Upper Confidence Bound ei Expected Improvement poi Probability of Improvement
kappa	tunable parameter kappa of GP Upper Confidence Bound, to balance exploitation against exploration, increasing kappa will make the optimized hyperparameters pursuing exploration.
eps	tunable parameter epsilon of Expected Improvement and Probability of Improvement, to balance exploitation against exploration, increasing epsilon will make the optimized hyperparameters are more spread out across the whole range.
optkernel	Kernel (aka correlation function) for the underlying Gaussian Process. This parameter should be a list that specifies the type of correlation function along with the smoothness parameter. Popular choices are square exponential (default) or matern 5/2
classes	set the number of classes. To use only with multiclass objectives.

Value

The test accuracy and a list of Bayesian Optimization result is returned:

• Best_Par a named vector of the best hyperparameter set found

• Best_Value the value of metrics achieved by the best hyperparameter set

• History a data.table of the bayesian optimization history

• Pred a data.table with validation/cross-validation prediction for each round of bayesian optimization history

Examples

# NOT RUN {
library(MlBayesOpt)

set.seed(71)
res0 <- xgb_opt(train_data = fashion_train,
                train_label = y,
                test_data = fashion_test,
                test_label = y,
                objectfun = "multi:softmax",
                evalmetric = "merror",
                classes = 10,
                init_points = 3,
                n_iter = 1)
# }