Fit a Quantile Regression Forest for Active Learning

Fits a Quantile Regression Forest (Meinshausen 2006) on the currently labeled soil dataset, using ranger::ranger(..., quantreg = TRUE). The resulting object carries the data, model, and an initial history entry and is the value on which al_query(), al_update() and al_loop() operate.

Usage

al_fit(labeled, target, covariates, coords = NULL, num.trees = 500L, ...)

Arguments

labeled: Data frame with observed target + covariates (+ optional coordinates).
target: Character, name of the target column.
covariates: Character vector of covariate column names.
coords: Optional length-2 character vector naming the x/y coordinate columns. Required by the "cost" query strategy.
num.trees: Integer, number of trees (default 500).
...: Additional arguments forwarded to ranger::ranger().

Value

A edaphos_al_model object.

References

Meinshausen N (2006). Quantile Regression Forests. Journal of Machine Learning Research 7, 983-999.

Examples

# \donttest{
  if (requireNamespace("sp", quietly = TRUE)) {
    data(meuse, package = "sp")
    m <- al_fit(
      labeled    = stats::na.omit(meuse[1:30, ]),
      target     = "lead",
      covariates = c("dist", "elev"),
      coords     = c("x", "y")
    )
    m
  }
#> <edaphos_al_model>
#>   target     : lead 
#>   covariates : dist, elev 
#>   coords     : x, y 
#>   n labeled  : 29 
#>   iterations : 0 
#>   last RMSE  : 41.99 
# }