pub fn time_series_split(
data: &Dataset,
n_splits: usize,
) -> Vec<(Dataset, Dataset)>Expand description
Time-series cross-validation with expanding training window.
Produces n_splits folds where each test set immediately follows its
training set. Data order is preserved — no shuffling.
For n_splits folds and n samples, each test chunk has size
n / (n_splits + 1). Fold i trains on [0 .. (i+1)*chunk] and tests
on [(i+1)*chunk .. (i+2)*chunk].