Bayesian Additive Regression Trees Model
Builds a BART model for regression or classification.
BARTMachineModel( num_trees = 50, num_burn = 250, num_iter = 1000, alpha = 0.95, beta = 2, k = 2, q = 0.9, nu = 3, mh_prob_steps = c(2.5, 2.5, 4)/9, verbose = FALSE, ... )
num_trees |
number of trees to be grown in the sum-of-trees model. |
num_burn |
number of MCMC samples to be discarded as "burn-in". |
num_iter |
number of MCMC samples to draw from the posterior distribution. |
alpha, beta |
base and power hyperparameters in tree prior for whether a node is nonterminal or not. |
k |
regression prior probability that E(Y|X) is contained in the interval (y_{min}, y_{max}), based on a normal distribution. |
q |
quantile of the prior on the error variance at which the data-based estimate is placed. |
nu |
regression degrees of freedom for the inverse sigma^2 prior. |
mh_prob_steps |
vector of prior probabilities for proposing changes to the tree structures: (GROW, PRUNE, CHANGE). |
verbose |
logical indicating whether to print progress information about the algorithm. |
... |
additional arguments to |
binary factor, numeric
alpha, beta, k, nu
Further model details can be found in the source link below.
In calls to varimp for BARTMachineModel, argument
metric may be specified as "splits" (default) for the
proportion of time each predictor is chosen for a splitting rule or as
"trees" for the proportion of times each predictor appears in a tree.
Argument num_replicates is also available to control the number of
BART replicates used in estimating the inclusion proportions [default: 5].
Variable importance is automatically scaled to range from 0 to 100. To
obtain unscaled importance values, set scale = FALSE. See example
below.
MLModel class object.
## Requires prior installation of suggested package bartMachine to run model_fit <- fit(sale_amount ~ ., data = ICHomes, model = BARTMachineModel) varimp(model_fit, metric = "splits", num_replicates = 20, scale = FALSE)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.