Model Validation Engine
Comprehensive model validation for IFRS 9 compliance - Advanced ML validation with cross-validation techniques and performance monitoring
No validations in progress
Upload Training Data
Drag & drop or click to upload CSV, Excel, or Text file with historical validation data
Data Preview (First 10 Rows)
| Sample_ID | y_true | y_pred | Features... |
|---|
Model Validation Techniques
Train-Test Split
Simple holdout method dividing data into training and test sets. Standard approach for initial model assessment.
Performance Metrics
| Metric | Value | Status |
|---|---|---|
| Accuracy | 92.5% | ✓ Pass |
| Precision | 89.3% | ✓ Pass |
| Recall | 85.1% | ⚠ Monitor |
| F1-Score | 87.1% | ✓ Pass |
K-Fold Cross-Validation
Robust evaluation method splitting data into k folds, rotating training and testing across all subsets.
Cross-Validation Results
| Metric | Mean | Std Dev | Status |
|---|---|---|---|
| Accuracy | 93.2% | ±2.1% | ✓ Pass |
| AUC-ROC | 0.91 | ±0.03 | ✓ Pass |
| MSE | 0.045 | ±0.012 | ⚠ Monitor |
Stratified K-Fold
Preserves class distribution in each fold. Essential for imbalanced classification datasets.
Stratified Results
| Metric | Class 0 | Class 1 | Status |
|---|---|---|---|
| Accuracy | 94.1% | 91.8% | ✓ Pass |
| F1-Score | 0.93 | 0.89 | ✓ Pass |
| Balanced Acc. | 92.9% | - | ✓ Pass |
Leave-One-Out (LOO)
Extreme form of cross-validation. Each sample used once as test set. High computational cost.
LOO Performance
| Metric | Value | Status |
|---|---|---|
| CV Accuracy | 94.8% | ✓ Pass |
| Estimate Variance | High | ⚠ Note |
| Computational Cost | Very High | ⚠ Consider |
Time Series Split
Chronological validation respecting temporal order. No randomization for time-dependent data.
Time Series Performance
| Period | MSE | R² | Status |
|---|---|---|---|
| Q1 2024 | 0.032 | 0.89 | ✓ Pass |
| Q2 2024 | 0.041 | 0.85 | ⚠ Monitor |
| Q3 2024 | 0.029 | 0.91 | ✓ Pass |
Nested Cross-Validation
Inner loop for hyperparameter tuning, outer loop for performance estimation. Most robust validation.
Nested CV Results
| Metric | Outer CV | Tuning | Status |
|---|---|---|---|
| Final Accuracy | 95.2% | 96.1% | ✓ Pass |
| AUC-ROC | 0.94 | 0.95 | ✓ Pass |
| Variance | Low | - | ✓ Stable |
Model Comparison Summary
| Technique | Accuracy | AUC | Drift Score | Status |
|---|---|---|---|---|
| Train-Test Split | 92.5% | 0.89 | 0.05 | ✅ Pass |
| K-Fold CV | 93.2% | 0.91 | 0.04 | ✅ Pass |
| Stratified K-Fold | 94.1% | 0.92 | 0.03 | ✅ Pass |
| Leave-One-Out | 94.8% | 0.93 | 0.02 | ✅ Pass |
| Time Series Split | 91.8% | 0.88 | 0.06 | ⚠️ Check |
| Nested CV | 95.2% | 0.94 | 0.02 | ✅ Pass |