Cross-Validation Simulator

Visualize k-fold, stratified k-fold, and LOOCV to understand proper model validation

Select Cross-Validation Method

Holdout

Single train/test split

k-Fold CV

k equal-sized folds

Stratified k-Fold

Preserves class distribution

LOOCV

Leave-one-out CV

Dataset Configuration

Number of Samples: 50

Number of Folds (k): 5

Class Imbalance: Balanced

Controls

Data Split Visualization

Training Data

Validation Data

Class 0

Class 1

Cross-Validation Results

Mean Accuracy

Std Deviation

Min Accuracy

Max Accuracy

Folds Completed

Total Samples

Fold-by-Fold Results

About Cross-Validation

Cross-validation is a resampling technique used to evaluate machine learning models on limited data. It provides a more reliable estimate of model performance than a single train/test split.

Holdout: Simple single split. Fast but high variance in estimate.
k-Fold CV: Data divided into k folds. Each fold used as validation once. Standard approach.
Stratified k-Fold: Like k-fold but maintains class distribution in each fold. Best for classification.
LOOCV: Each sample used as validation once. Maximum data use but very slow.

Tips: Watch how different methods split the data. Notice how stratified k-fold maintains class balance in each fold, crucial for imbalanced datasets. Compare variance across folds to see reliability.