wt_xgboost: wt_xgboost

In MoritzFeigl/wateRtemp: Water temperature prediction modelling

Description

XGBoost implementation for stream water temperature prediction including Bayesian hyperparameter optimization. All results are stored automatically in the folder catchment/model_name.

Usage

wt_xgboost(
  train_data,
  test_data = NULL,
  catchment = NULL,
  cv_mode = "repCV",
  model_name = NULL,
  no_cores = parallel::detectCores() - 1,
  seed = NULL,
  n_iter = 40,
  n_random_initial_points = 20
)

Arguments

train_data	Data frame containing training data created by using wt_preprocessing()
test_data	Data frame containing test data created by using wt_preprocessing()
catchment	Catchment name as string, used for storing results in current working directory.
cv_mode	Cross-validation mode. Can either be "repCV" for a 5times repeated 10-fold CV or "timeseriesCV" for a timeslice CV using intial window=730, horizon=90 and skip=60.
model_name	Name of this particular model run as string, used for storing results in the catchment folder.
no_cores	Number of cores used for computation. If NULL parallel::detectCores() - 1 is applied.
seed	Random seed.
n_iter	Number of iteration steps for bayesian hyperparameter optimization.
n_random_initial_points	Number of sampled initial random points for bayesian hyperparameter optimization

Examples

## Not run: 
data(test_catchment)
wt_preprocess(test_catchment)
train_data <- feather::read_feather("test_catchment/train_data.feather")
test_data <- feather::read_feather("test_catchment/test_data.feather")

wt_xgboost(train_data, test_data, "test_catchment", "repCV", "standard_xgboost")

## End(Not run)

MoritzFeigl/wateRtemp documentation built on Sept. 6, 2021, 6:58 a.m.