MQR_gbm: Multiple Quantile Regression Using Gradient Boosted Decision...
In jbrowell/ProbCast: Flexible Probabilistic Forecasting

MQR_gbm

R Documentation

Multiple Quantile Regression Using Gradient Boosted Decision Trees (depreciated)

Description

This function is now depreciated and may be removed in future versions of this package. Use qreg_gbm() instead.

Usage

MQR_gbm(
  data,
  formula,
  quantiles = c(0.25, 0.5, 0.75),
  CVfolds = NULL,
  gbm_params = list(...),
  perf.plot = F,
  parallel = F,
  pred_ntree = NULL,
  cores = NULL,
  pckgs = NULL,
  para_over_q = FALSE,
  Sort = T,
  SortLimits = NULL
)

Arguments

`data`	A `data.frame` containing target and explanatory variables. May optionally contain a collumn called "kfold" with numbered/labeled folds and "Test" for test data.
`quantiles`	The quantiles to fit models for.
`CVfolds`	Control for cross-validation if not supplied in `data`.
`gbm_params`	List of parameters to be passed to `fit.gbm()`.
`perf.plot`	Plot GBM performance?
`parallel`	`boolean` parallelize cross-validation process? Parallelisation is over cross-validation folds by default, optionally over quantiles by setting codepara_over_q=T.
`pred_ntree`	predict using a user-specified tree. If unspecified an out-of-the bag estimate will be used unless interval gbm cross-validation folds are specified in `gbm_params`.
`cores`	if `parallel` is TRUE then number of available cores
`pckgs`	if `parallel` is TRUE then specify packages required for each worker (e.g. c("data.table) if data stored as such).
`para_over_q`	if `parallel` is TRUE then paralellize over quantiles? Defalts to FALSE i.e."kfold".
`Sort`	`boolean` Sort quantiles using `SortQuantiles()`?
`SortLimits`	`Limits` argument to be passed to `SortQuantiles()`. Constrains quantiles to upper and lower limits given by `list(U=upperlim,L=lowerlim)`.
`formala`	A `formula` object with the response on the left of an ~ operator, and the terms, separated by + operators, on the right

Details

The returned predictive quantiles are those produced out-of-sample for each cross-validation fold (using models trained on the remaining folds but not "Test" data). Predictive quantiles corresponding to "Test" data are produced using models trained on all non-test data.