Create a regressor object for Probabilistic Machine Learning

regressor(
  x,
  y,
  model = c("ranger", "extratrees", "ridge", "bcn", "glmnet", "krr", "xgboost", "svm"),
  pi_method = c("none", "splitconformal", "jackknifeplus", "kdesplitconformal",
    "bootsplitconformal", "kdejackknifeplus", "bootjackknifeplus", "surrsplitconformal",
    "surrjackknifeplus"),
  level = 95,
  B = 100,
  nb_hidden = 0,
  nodes_sim = c("sobol", "halton", "unif"),
  activ = c("relu", "sigmoid", "tanh", "leakyrelu", "elu", "linear"),
  engine = NULL,
  params = NULL,
  type_split = c("stratify", "sequential"),
  seed = 123
)

Arguments

x: Input matrix or data frame of features
y: Vector of target values
model: Model to use for regression
pi_method: Method to use for conformal prediction
level: Confidence level for conformal prediction
B: Number of simulations for conformal prediction
nb_hidden: Number of nodes in the hidden layer
nodes_sim: Type of simulations for hidden nodes
activ: Activation function for hidden layer
engine: Engine to use for fitting the model
params: Additional parameters passed to the model
type_split: Type of data splitting for split conformal prediction: "stratify" (for classical supervised learning) "sequential" (when the data sequential ordering matters)
seed: Reproducibility seed for randomization
...: Additional arguments passed to Regressor$new()

Value

A regressor object of class "regressor"

Examples


X <- mtcars[, -1]
y <- mtcars$mpg
X_train <- X[1:25, ]
y_train <- y[1:25]
X_test <- X[26:32, ]
y_test <- y[26:32]
reg <- regressor(X_train, y_train, pi_method = "splitconformal", level = 95)

print(sqrt(mean((y_test - predict(reg, newx = X_test))^2)))
#> Error in h(simpleError(msg, call)): error in evaluating the argument 'x' in selecting a method for function 'print': error in evaluating the argument 'x' in selecting a method for function 'mean': non-numeric argument to binary operator