GLMTreeHyperparameters

Properties

regressors

array | null≥ 1 items

Features entering the linear model in each leaf. NULL uses every feature; an empty model is not expressible here, since a tree of intercepts is a decision tree and CART fits one. Values must name training features.

partitioning_variables

array | null≥ 1 items

Features the tree may split on. NULL uses every feature. Values must name training features.

alpha

number | object≤ 1> 0

Significance level a parameter-instability test must reach before a node is split. The one setting that governs tree size.

bonferroni

boolean

Bonferroni-adjust the test p-values for the number of partitioning variables.

minsize

integer | object | null≥ 1

Fewest cases a node may hold. NULL uses ten times the number of parameters in a leaf's model, which scales with the number of regressors and can forbid splitting entirely on a wide design. Cannot exceed the number of cases in the training data.

maxdepth

integer | object | null≥ 1

Deepest the tree may grow. NULL is unlimited.

mtry

integer | object | null≥ 1

Partitioning variables sampled at each node. NULL tests every one. Cannot exceed the number of features in the training data.

trim

number | object≥ 0

Trimming for the split-point search: a fraction of cases below 1, a count at or above it.

breakties

boolean

Break ties in numeric partitioning variables by adding noise.

prune

string | object | null

Post-prune the grown tree by an information criterion. NULL keeps every split the tests admitted.

one of"AIC""BIC"

restart

boolean

Restart each leaf's model from scratch rather than from its parent's fit.

dfsplit

boolean

Count each split against the degrees of freedom when pruning.

numsplit

string | object

Where a numeric split is placed: at the last value of the left branch, or midway between it and the next.

one of"left""center"

catsplit

string | object

How a categorical variable is split: into two groups of levels, or one branch per level.

one of"binary""multiway"

ordinal

string | object

Test statistic for an ordered partitioning variable.

one of"chisq""max""L2"

vcov

string

Covariance estimator the instability test is built on.

one of"opg""info""sandwich"

nrep

integer≥ 1

Simulation replicates for the p-value of an ordinal split statistic.

ifw

boolean | object

Inverse Frequency Weighting in classification.

Relationships