xgboost evaluation metric wrapper

The wrapper works only if both the wrapper and the original evaluation metric are existing. Requires Matrix and xgboost packages.

xgb.wrap.metric(f, name)

Arguments

f	Type: function. The function to wrap from xgboost. Requires the following order of arguments for the function to work: preds, labels, and returns a single value.
name	Type: character. The evaluation metric name which is printed and used for xgboost variables.

Value

The wrapping function.

Examples

# Note: this example unexpectedly fails when using pkgdown.

library(xgboost)
library(Matrix)
data(agaricus.train, package = "xgboost")
data(agaricus.test, package = "xgboost")

dtrain <- xgboost::xgb.DMatrix(agaricus.train$data, label = agaricus.train$label)
dtest <- xgboost::xgb.DMatrix(agaricus.test$data, label = agaricus.test$label)
watchlist <- list(train = dtrain, eval = dtest)

evalerror <- function(preds, labels) {
  return(mean(abs(preds - labels)))
}
evalerror_wrap <- xgb.wrap.metric(f = evalerror, "mae")

param <- list(max_depth = 2, eta = 1, silent = 1, nthread = 1,
              objective = "binary:logistic", eval_metric = evalerror_wrap)
bst <- xgboost::xgb.train(param, dtrain, nrounds = 2, watchlist)
#> Error in evalerror(preds, xgboost::getinfo(dtrain, "label")): could not find function "evalerror"