Struct StratifiedKFold 

pub struct StratifiedKFold<L> { /* private fields */ }

Stratified K-fold cross-validation.

Each fold keeps roughly the same class distribution as the full dataset. This ports the previously hand-rolled stratification logic from the guide’s Evaluation addendum into a real, tested implementation: samples are grouped by class label, then each class’s samples are distributed across the folds so every fold receives a proportional slice of every class.

The label type is generic (L: Eq + Hash + Clone), so string, integer or enum labels all work with no forced mapping step — the same convention as the sibling imbalance-rs crate. Labels are supplied at construction and stored, which is what lets a stratified splitter satisfy the plain CvSplitter interface (see the module docs).

Small classes

A class with fewer than n_splits samples cannot appear in every fold. Such a class is spread across as many folds as it can fill (one sample each, leading folds first) and a warning is emitted on stderr — matching the “warn and adjust rather than hard-error where reasonable” policy.

use ndarray::array;
use model_selection_rs::splitters::{CvSplitter, StratifiedKFold};

let y = array![0, 0, 0, 0, 1, 1, 1, 1];
let skf = StratifiedKFold::new(2, &y).unwrap();
let splits = skf.split(y.len()).unwrap();
// Each fold's test set holds two 0s and two 1s.
assert_eq!(splits.len(), 2);

Implementations

impl<L: Eq + Hash + Clone> StratifiedKFold<L>

pub fn new(n_splits: usize, y: &Array1<L>) -> Result<Self>

Create a StratifiedKFold over the class labels y.

Errors

Returns ModelSelectionError::InvalidSplitCount if n_splits < 2.

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

Shuffle each class’s samples (deterministically, from seed) before distributing them across folds.

Trait Implementations

impl<L: Clone> Clone for StratifiedKFold<L>

fn clone(&self) -> StratifiedKFold<L>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<L: Eq + Hash + Clone> CvSplitter for StratifiedKFold<L>

fn split(&self, n_samples: usize) -> Result<Vec<(Vec<usize>, Vec<usize>)>>

Produce every (train_indices, test_indices) pair for n_samples rows.

fn n_splits(&self) -> usize

Number of splits this strategy yields.

impl<L: Debug> Debug for StratifiedKFold<L>

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

Formats the value using the given formatter.