Struct ConTree 

pub struct ConTree {
    pub tree: Tree,
    /* private fields */
}

Exact ConTree search for optimal decision trees on continuous features.

A depth-first branch-and-bound search with caching. For each feature it explores candidate thresholds by bisection over intervals, and uses the errors already found to prune whole intervals of thresholds at once.

Fields

tree: Tree

Implementations

impl ConTree

pub fn with_config(config: SearchConfig) -> Self

Builds a solver from a SearchConfig.

This is the preferred constructor; Self::new takes the same settings as positional arguments.

pub fn new( min_sup: usize, max_depth: usize, max_time: f64, max_error: usize, split_selection_strategy: PointSelector, max_gap: usize, use_heuristic: bool, fast_d2: bool, ) -> Self

Builds a solver from positional settings. See Self::with_config.

pub fn with_random_state(self, seed: u64) -> Self

Seeds the random generator used by PointSelector::Random, making runs reproducible. Unseeded solvers draw their seed from the OS.

pub fn fit(&mut self, dataset: &Dataset) -> Result<FitOutcome, SearchError>

Searches for the optimal tree.

Returns the tree, the search counters, and why the search stopped. Only SearchStatus::Optimal means the tree is proven best for the given depth and minimum support.

pub fn trajectory(&self) -> &[(f64, usize)]

The anytime profile of the last fit: (seconds, error) at every improvement of the root incumbent.

pub fn status(&self) -> SearchStatus

Why the last fit stopped.

pub fn statistics(&self) -> Statistics

pub fn select_point(&mut self, config: &SearchConfig, bound: &Bound) -> usize

pub fn get_solution_tree(&mut self)

Rebuilds self.tree from the cache.