CrossValidationStrategy
Enums · AiDotNet.Evaluation.Enums
Specifies the cross-validation strategy to use for model evaluation.
For Beginners
Cross-validation is like giving your model multiple “practice tests” on different portions of your data. Each strategy splits the data differently:
- KFold: Divides data into K equal parts, trains on K-1, tests on 1, repeats K times
- Stratified: Same as KFold but preserves class proportions (important for imbalanced data)
- TimeSeries: Respects time order - always trains on past, tests on future
- Group: Keeps related samples together (e.g., all data from one patient stays together)
Choose based on your data type and whether you have class imbalance, time dependencies, or grouped samples.
How It Works
Cross-validation is a technique for assessing how well a model generalizes to independent data. Different strategies are appropriate for different data types and problem domains.
Fields
| Field | Summary |
|---|---|
CrossValidationStrategy AdversarialSplit |
Adversarial Split deliberately creates challenging test sets to evaluate robustness. |
CrossValidationStrategy BlockedTimeSeriesSplit |
Blocked Time Series Split adds gaps between train and test to prevent information leakage. |
CrossValidationStrategy BootstrapCV |
Bootstrap CV uses bootstrap resampling (sampling with replacement). |
CrossValidationStrategy CombinatorialPurgedCV |
Combinatorial Purged CV generates all possible train/test combinations with purging. |
CrossValidationStrategy GroupKFold |
Group K-Fold ensures samples from the same group never appear in both train and test. |
CrossValidationStrategy GroupShuffleSplit |
Group Shuffle-Split randomly samples groups for train/test while respecting group boundaries. |
CrossValidationStrategy KFold |
Standard K-Fold cross-validation. |
CrossValidationStrategy LeaveOneOut |
Leave-One-Out (LOO) uses each sample as a single test case while training on all others. |
CrossValidationStrategy LeavePOut |
Leave-P-Out uses all possible combinations of P samples as test sets. |
CrossValidationStrategy MonteCarloCV |
Monte Carlo CV (repeated random sub-sampling) creates random train/test splits. |
CrossValidationStrategy NestedCV |
Nested CV uses an outer loop for performance estimation and inner loop for hyperparameter tuning. |
CrossValidationStrategy PurgedKFold |
Purged K-Fold removes samples from training that are temporally close to test samples. |
CrossValidationStrategy RepeatedKFold |
Repeated K-Fold runs standard K-Fold multiple times with different random shuffles. |
CrossValidationStrategy RepeatedStratifiedKFold |
Repeated Stratified K-Fold combines stratification with repetition. |
CrossValidationStrategy ShuffleSplit |
Shuffle-Split creates random train/test splits with configurable sizes. |
CrossValidationStrategy SlidingWindowSplit |
Sliding Window Split uses fixed-size training windows that slide through time. |
CrossValidationStrategy StratifiedKFold |
Stratified K-Fold preserves the percentage of samples for each class in each fold. |
CrossValidationStrategy StratifiedShuffleSplit |
Stratified Shuffle-Split maintains class proportions in random splits. |
CrossValidationStrategy TimeSeriesSplit |
Time Series Split uses expanding training windows. |
int value__ |