Skip to contents

Fit a multivariate regression forest

Source: R/fit_mv_forest.R
fit_mv_forest.Rd

Drop-in replacement for fit_forest() when type = "regression". Returns an object with the same interface ($forest.wt, $proximity, $membership, $xvar, $yvar, $ntree, $xvar.names) so downstream code in multiRF works without changes.

Usage

fit_mv_forest(
  X,
  Y,
  ntree = 500L,
  mtry = NULL,
  ytry = NULL,
  nsplit = 10L,
  forest.wt = "all",
  proximity = c("all", "inbag", "oob", "none"),
  nodesize = 5L,
  max_depth = 0L,
  seed = 529L,
  samptype = c("swor", "swr"),
  nthread = getOption("multiRF.nthread", 0L),
  xvar.wt = NULL,
  yvar.wt = NULL,
  enhanced_prox = FALSE,
  sibling_gamma = 0.5,
  leaf_embed_dim = 10L
)

Arguments

X

Data frame or matrix of predictor features (n x px).

Y

Data frame or matrix of response features (n x qy).

ntree

Number of trees.

mtry

Number of candidate X variables per split. Accepts an integer, a formula string ("sqrt(p)", "p/3", "p/2"), or NULL for the default (ceiling(px/3) for regression, ceiling(sqrt(px)) for classification). In formulas, p is the number of predictor columns.

ytry

Number of candidate Y variables per split. Accepts an integer, a formula string ("sqrt(p)", "p/3"), or NULL for the default (ceiling(qy/3)). In formulas, p is the number of Y columns.

nsplit

Number of candidate numeric cutpoints per split variable.

forest.wt

Forest-weight mode: "all", "inbag", or "oob".

proximity

Proximity output mode.

nodesize

Minimum terminal node size. Default: 5.

max_depth

Maximum tree depth (0 = unlimited). Default: 0.

seed

Random seed.

samptype

Sampling scheme: "swor" or "swr".

nthread

Number of threads used for fitting.

xvar.wt

Optional non-negative predictor sampling weights.

yvar.wt

Optional non-negative response sampling weights.

enhanced_prox

Logical; whether to compute enhanced proximity.

sibling_gamma

Strength of the sibling-leaf correction used by enhanced proximity.

leaf_embed_dim

Embedding dimension used by enhanced proximity.

Value

A list compatible with rfsrc output, containing:

forest.wt

n x n forest weight matrix

proximity

n x n proximity matrix

membership

n x ntree terminal node membership

xvar

predictor data frame

yvar

response data frame

xvar.names

character vector of predictor names

ntree

number of trees

tree_info

per-tree structure for IMD