Regressor.RdThe `Regressor` class contains supervised regression models
This class implements models:
Linear model
see https://www.researchgate.net/publication/380760578_Boosted_Configuration_neural_Networks_for_supervised_classification
Extremely Randomized Trees; see https://link.springer.com/article/10.1007/s10994-006-6226-1
Elastic Net Regression; see https://glmnet.stanford.edu/
Kernel Ridge Regression; see for example https://www.jstatsoft.org/article/view/v079i03
Random Forest; see https://www.jstatsoft.org/article/view/v077i01
Ridge regression; see https://arxiv.org/pdf/1509.09169
a scalable tree boosting system see https://arxiv.org/abs/1603.02754
Support Vector Machines, see https://cran.r-project.org/web/packages/e1071/vignettes/svmdoc.pdf
Random Vector Functional Network, see https://www.researchgate.net/publication/332292006_Online_Bayesian_Quasi-Random_functional_link_networks_application_to_the_optimization_of_black_box_functions
learningmachine::Base -> Regressor
namename of the class
typetype of supervised learning method implemented
modelfitted model
methodsupervised learning method in c('lm', 'ranger', 'extratrees', 'ridge', 'bcn', 'glmnet', 'krr', 'xgboost', 'svm')
X_traintraining set features; do not modify by hand
y_traintraining set response; do not modify by hand
pi_methodtype of prediction interval in c("splitconformal", "kdesplitconformal", "bootsplitconformal", "jackknifeplus", "kdejackknifeplus", "bootjackknifeplus", "surrsplitconformal", "surrjackknifeplus")
levelan integer; the level of confidence (default is 95, for 95 per cent) for prediction intervals
Ban integer; the number of simulations when 'level' is not NULL
nb_hiddennumber of nodes in the hidden layer, for construction of a quasi- randomized network
nodes_simtype of 'simulations' for hidden nodes, if nb_hidden > 0;
takes values in c("sobol", "halton", "unif")
activactivation function's name for the hidden layer, in the construction of a quasi-randomized network; takes values in c("relu", "sigmoid", "tanh", " leakyrelu", "elu", "linear")
enginecontains fit and predic lower-level methods for the given method;
do not modify by hand
paramsadditional parameters passed to method when calling fit
do not modify by hand
seedan integer; reproducibility seed for methods that include randomization
Inherited methods
learningmachine::Base$get_B()learningmachine::Base$get_activ()learningmachine::Base$get_engine()learningmachine::Base$get_level()learningmachine::Base$get_method()learningmachine::Base$get_model()learningmachine::Base$get_name()learningmachine::Base$get_nb_hidden()learningmachine::Base$get_nodes_sim()learningmachine::Base$get_params()learningmachine::Base$get_pi_method()learningmachine::Base$get_seed()learningmachine::Base$get_type()learningmachine::Base$set_B()learningmachine::Base$set_activ()learningmachine::Base$set_engine()learningmachine::Base$set_level()learningmachine::Base$set_method()learningmachine::Base$set_model()learningmachine::Base$set_nb_hidden()learningmachine::Base$set_nodes_sim()learningmachine::Base$set_pi_method()learningmachine::Base$set_seed()learningmachine::Base$summary()
new()Create a new object.
Regressor$new(
name = "Regressor",
type = "regression",
model = NULL,
method = NULL,
X_train = NULL,
y_train = NULL,
pi_method = c("none", "splitconformal", "jackknifeplus", "kdesplitconformal",
"bootsplitconformal", "kdejackknifeplus", "bootjackknifeplus", "surrsplitconformal",
"surrjackknifeplus"),
level = 95,
B = 100,
nb_hidden = 0,
nodes_sim = c("sobol", "halton", "unif"),
activ = c("relu", "sigmoid", "tanh", "leakyrelu", "elu", "linear"),
engine = NULL,
params = NULL,
seed = 123
)fit()Fit model to training set
Regressor$fit(X, y, type_split = c("stratify", "sequential"), ...)Xa matrix of covariates (i.e explanatory variables)
ya vector, the response (i.e variable to be explained)
type_splittype of data splitting for split conformal prediction: "stratify" (for classical supervised learning) "sequential" (when the data sequential ordering matters)
...additional parameters to learning algorithm (see vignettes)
predict()Predict model on test set
fit_predict()Fit model to training set and predict on test set
Regressor$fit_predict(
X,
y,
pct_train = 0.8,
score = ifelse(is.factor(y), yes = function(preds, y_test) mean(preds == y_test), no =
function(preds, y_test) sqrt(mean((preds - y_test)^2))),
level = NULL,
pi_method = c("none", "splitconformal", "jackknifeplus", "kdesplitconformal",
"bootsplitconformal", "kdejackknifeplus", "bootjackknifeplus", "surrsplitconformal",
"surrjackknifeplus"),
B = 100,
seed = 123,
graph = FALSE,
...
)update()update model in an online fashion (for now, only implemented for 'rvfl' models")