Struct DecisionTreeClassifier 

pub struct DecisionTreeClassifier<F> {
    pub max_depth: Option<usize>,
    pub min_samples_split: usize,
    pub min_samples_leaf: usize,
    pub min_weight_fraction_leaf: F,
    pub min_impurity_decrease: F,
    pub ccp_alpha: F,
    pub max_leaf_nodes: Option<usize>,
    pub criterion: ClassificationCriterion,
    pub class_weight: ClassWeight<F>,
    /* private fields */
}

CART decision tree classifier.

Builds a binary tree by recursively finding the feature and threshold that maximises the reduction in the chosen impurity criterion (Gini or Entropy).

Type Parameters

• F: The floating-point type (f32 or f64).

Fields

max_depth: Option<usize>

Maximum depth of the tree. None means unlimited.

min_samples_split: usize

Minimum number of samples required to split an internal node.

min_samples_leaf: usize

Minimum number of samples required in a leaf node.

min_weight_fraction_leaf: F

Minimum weighted fraction of the total sample weight required at a leaf.

Mirrors sklearn’s min_weight_fraction_leaf (_classes.py:946, default 0.0). For uniform sample weights the effective minimum leaf size becomes max(min_samples_leaf, ceil(min_weight_fraction_leaf · N)) where N is the total training-sample count (_classes.py:371, _splitter.pyx:470).

min_impurity_decrease: F

Minimum tree-normalized weighted impurity decrease required to split a node.

Mirrors sklearn’s min_impurity_decrease (_classes.py:946, default 0.0). A node is made a leaf when the split’s improvement N_t/N·(parent − N_tL/N_t·imp_L − N_tR/N_t·imp_R) satisfies improvement + EPSILON < min_impurity_decrease (_tree.pyx:284).

ccp_alpha: F

Complexity parameter for Minimal Cost-Complexity Pruning.

Mirrors sklearn’s ccp_alpha (_classes.py:946, default 0.0, Interval(Real, 0.0, None, closed="left"), _classes.py:123). After the tree is grown, the subtree with the largest cost complexity that is smaller than ccp_alpha is chosen (Breiman weakest-link pruning, _tree.pyx::_cost_complexity_prune). 0.0 ⇒ no pruning.

max_leaf_nodes: Option<usize>

Maximum number of leaf nodes. None ⇒ unlimited (depth-first growth).

Mirrors sklearn’s max_leaf_nodes (_classes.py:946, default None, Interval(Integral, 2, None, closed="left"), _classes.py:121). When Some(k), the tree is grown best-first (highest impurity improvement expanded first, BestFirstTreeBuilder, _tree.pyx:407) until it has k leaves (2k−1 nodes) or no expandable frontier node remains. None keeps the byte-identical depth-first build.

criterion: ClassificationCriterion

Splitting criterion.

class_weight: ClassWeight<F>

Per-class weighting (sklearn class_weight, _classes.py:801, default None). Expanded to per-sample weights at fit and folded into every weighted quantity (node class counts, gini/entropy, the leaf distribution / predict_proba, the min_weight_fraction_leaf gate). ClassWeight::None (default) ⇒ a byte-identical unweighted tree.

Implementations

impl<F: Float> DecisionTreeClassifier<F>

pub fn new() -> Self

Create a new DecisionTreeClassifier with default settings.

Defaults: max_depth = None, min_samples_split = 2, min_samples_leaf = 1, min_weight_fraction_leaf = 0.0, min_impurity_decrease = 0.0, ccp_alpha = 0.0, max_leaf_nodes = None, criterion = Gini, class_weight = ClassWeight::None.

pub fn with_max_depth(self, max_depth: Option<usize>) -> Self

Set the maximum tree depth.

pub fn with_min_samples_split(self, min_samples_split: usize) -> Self

Set the minimum number of samples required to split a node.

pub fn with_min_samples_leaf(self, min_samples_leaf: usize) -> Self

Set the minimum number of samples required in a leaf node.

pub fn with_min_weight_fraction_leaf(self, min_weight_fraction_leaf: F) -> Self

Set the minimum weighted fraction of the total sample weight required at a leaf (sklearn min_weight_fraction_leaf, _classes.py:946).

pub fn with_min_impurity_decrease(self, min_impurity_decrease: F) -> Self

Set the minimum tree-normalized weighted impurity decrease required to split a node (sklearn min_impurity_decrease, _classes.py:946).

pub fn with_ccp_alpha(self, ccp_alpha: F) -> Self

Set the complexity parameter for Minimal Cost-Complexity Pruning (sklearn ccp_alpha, _classes.py:946, default 0.0).

pub fn with_max_leaf_nodes(self, max_leaf_nodes: Option<usize>) -> Self

Set the maximum number of leaf nodes (sklearn max_leaf_nodes, _classes.py:946, default None). Some(k) switches the build to best-first growth (BestFirstTreeBuilder, _tree.pyx:407).

pub fn with_criterion(self, criterion: ClassificationCriterion) -> Self

Set the splitting criterion.

pub fn with_class_weight(self, class_weight: ClassWeight<F>) -> Self

Set the per-class weighting (sklearn class_weight, _classes.py:801). ClassWeight::None (default) leaves every class at 1.0; ClassWeight::Balanced uses n_samples / (n_classes · count_c); ClassWeight::Explicit takes a per-class-label weight map (compute_class_weight, class_weight.py:20). The weights are expanded to per-sample weights at fit and fold into every weighted node quantity.

Trait Implementations

impl<F: Clone> Clone for DecisionTreeClassifier<F>

fn clone(&self) -> DecisionTreeClassifier<F>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<F: Debug> Debug for DecisionTreeClassifier<F>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<F: Float> Default for DecisionTreeClassifier<F>

fn default() -> Self

Returns the “default value” for a type.

impl<'de, F> Deserialize<'de> for DecisionTreeClassifier<F>

where F: Deserialize<'de>,

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl<F: Float + Send + Sync + 'static> Fit<ArrayBase<OwnedRepr<F>, Dim<[usize; 2]>>, ArrayBase<OwnedRepr<usize>, Dim<[usize; 1]>>> for DecisionTreeClassifier<F>

fn fit( &self, x: &Array2<F>, y: &Array1<usize>, ) -> Result<FittedDecisionTreeClassifier<F>, FerroError>

Fit the decision tree classifier on the training data.

Errors

Returns FerroError::ShapeMismatch if x and y have different numbers of samples. Returns FerroError::InsufficientSamples if there are no samples. Returns FerroError::InvalidParameter if hyperparameters are invalid.

type Fitted = FittedDecisionTreeClassifier<F>

The fitted model type returned by fit.

type Error = FerroError

The error type returned by fit.

impl<F: Float + ToPrimitive + FromPrimitive + Send + Sync + 'static> PipelineEstimator<F> for DecisionTreeClassifier<F>

fn fit_pipeline( &self, x: &Array2<F>, y: &Array1<F>, ) -> Result<Box<dyn FittedPipelineEstimator<F>>, FerroError>

Fit this estimator on the given data.

impl<F> Serialize for DecisionTreeClassifier<F>

where F: Serialize,

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.