drf (version 1.1.0)

predict.drf: Predict with a drf forest

Description

Predict with a drf forest

Usage

# S3 method for drf
predict(
  object,
  newdata = NULL,
  transformation = NULL,
  functional = NULL,
  num.threads = NULL,
  custom.functional = function(y, w) apply(y, 2, sum(y * w)),
  ...
)

Arguments

object

The trained drf forest.

newdata

Points at which predictions should be made. If NULL, makes out-of-bag predictions on the training set instead (i.e., provides predictions at Xi using only trees that did not use the i-th training example). Note that this matrix (or vector) should have the number of columns as the training matrix, and that the columns must appear in the same order.

transformation

a function giving a transformation of the responses, by default if NULL, the identity function(y) y is used.

functional

which type of statistical functional. One option between:

• "mean"the conditional mean, the returned value is a list containing a matrix mean of size n x f, where n denotes the number of observation in newdata and f the dimension of the transformation.

• "sd"the conditional standard deviation, the returned value is a list containing a matrix sd of size n x f, where n denotes the number of observation in newdata and f the dimension of the transformation.

• "quantile"the conditional quantiles, the returned value is a list containing an array quantile of size n x f x q, where n denotes the number of observation in newdata, f the dimension of the transformation and q the number of desired quantiles.

• "cor"the conditional correlation, the returned value is a list containing an array cor of size n x f x f, where n denotes the number of observation in newdata, f the dimension of the transformation.

• "cov"the conditional covariance, the returned value is a list containing an array cor of size n x f x f, where n denotes the number of observation in newdata, f the dimension of the transformation.

• "normalPredictionScore"a prediction score based on an asymptotic normality assumption, the returned value is a list containing a list of functions normalPredictionScore of size n, where n denotes the number of observation in newdata. Here the transformation should be uni-dimensional.

• "custom"a custom function provided by the user, the returned value is a list containing a matrix custom of size n x f, where n denotes the number of observation in newdata and f the dimension of the output of the function custom.functional.

• "MQ"multivariate quantiles, return a list containing a matrix of the inputed ranks u (that should be provided as an argument of the predict function) along with a list of the different corresponding MQ (same size as u).

num.threads

Number of threads used in training. If set to NULL, the software automatically selects an appropriate amount.

custom.functional

a function giving the custom functional when functional is set to "custom". This should be a function f(y,w) using the training response matrix y and the weights w at a single testing point.

...

additional parameters.

Value

a list containing an entry with the same name as the functional selected.

Examples

# NOT RUN {
# Train a distributional random forest with CART splitting rule.
n <- 100
p <- 2
X <- matrix(rnorm(n * p), n, p)
Y <- X + matrix(rnorm(n * p), ncol=p)
drf.forest <- drf(X = X, Y = Y)

# Predict conditional correlation.
X.test <- matrix(0, 101, p)
X.test[, 1] <- seq(-2, 2, length.out = 101)
cor.pred <- predict(drf.forest, X.test, functional = "cor")

# Predict on out-of-bag training samples.
cor.oob.pred <- predict(drf.forest,  functional = "cor")

# Train a distributional random forest with "FourierMMD" splitting rule.
n <- 100
p <- 2
X <- matrix(rnorm(n * p), n, p)
Y <- X + matrix(rnorm(n * p), ncol=p)
drf.forest <- drf(X = X, Y = Y, splitting.rule = "FourierMMD", num.features = 10)

# Predict conditional correlation.
X.test <- matrix(0, 101, p)
X.test[, 1] <- seq(-2, 2, length.out = 101)
cor.pred <- predict(drf.forest, X.test, functional = "cor")

# Predict on out-of-bag training samples.
cor.oob.pred <- predict(drf.forest,  functional = "cor")


# }