tweakr: tweakr

In illoRocks/tweakr:

Description

Parametertuning

Usage

tweakr(train_set, params, func_train, k = 5, folds = NULL,
  save_path = NULL, save_freq = 10, twk_object = NULL, ...)

Arguments

train_set	Training data.
params	List of parameters.
func_train	Function to train a model. The arguments must be 'param', 'train' and 'test' and return the fitted object.
k	Number of folds (Default: 5).
folds	custom folds (Default: NULL).
save_path	The path where the model are stored (Default: NULL).
save_freq	The frequence of model saving. (Defaut: 10)
twk_object	tweakr object to continue training (Default: NULL).
...	Additional arguments for tweakr functions.

Details

Parametrtuning for custom models.

Examples

library(rpart)
set.seed(123)

twk <- tweakr(train_set = iris,
params = list(cp=c(.01,.05)),
k = 10,
func_train = function(param, train, test) {
  model <- rpart(Species~. , train, control = rpart.control(cp = param$cp))
  pred <- predict(model, test, type = "class")
  error <- sum(pred == test$Species) / nrow(test)
  list(fit=model, pred=pred, eval=error)
})

prediction <- predict(twk,
                      iris,
                      func_predict = function(fit, test) predict(fit, test, type = "prob"))

illoRocks/tweakr documentation built on July 23, 2019, 3:54 p.m.