cross_validation

Sampling, Validation and ML Diagnostics

Purged cross-validation utilities designed for label overlap and leakage control.

Why This Module Exists

Time-dependent labels violate IID assumptions; purging/embargoing reduces leakage bias.

Key Public APIs

ml_cross_val_score
ml_get_train_times
PurgedKFold
Scoring

Core Math

Purged Train Set

\[\mathcal{T}_{train}=\mathcal{T}\setminus(\mathcal{T}_{test}\oplus e)\]

Embargo

\[e=\lfloor p\cdot T\rfloor\]

Code Examples

Configure PurgedKFold

use openquant::cross_validation::PurgedKFold;

let cv = PurgedKFold::new(5, 0.01);

Implementation Notes

Always align event end-times when purging.
Report variance across folds, not only mean score.