## Bar Chart: Performance Metrics of Surrogate Models
### Overview
The chart compares three performance metrics (Kendal Tau Correlation, MSE AVM, and MSE 1-day STD-DEV) across four surrogate models (XGBoost, LGBoost, MLP, MLP Ensemble). Each model is represented by grouped bars in red (Kendal Tau), purple (MSE AVM), and cyan (MSE 1-day STD-DEV). Values are labeled on top of bars, with axes scaled to 0–1 for Kendal Tau and 0–15 for Mean Squared Error.
### Components/Axes
- **X-axis**: Surrogate Models (XGBoost, LGBoost, MLP, MLP Ensemble)
- **Left Y-axis**: Kendal Tau Correlation (0–1)
- **Right Y-axis**: Mean Squared Error (0–15)
- **Legend**: Located at bottom-left, mapping colors to metrics:
- Red: Kendal Tau Correlation
- Purple: MSE AVM
- Cyan: MSE 1-day STD-DEV
### Detailed Analysis
#### Kendal Tau Correlation (Red Bars)
- **XGBoost**: 0.98 (highest)
- **LGBoost**: 0.94
- **MLP**: 0.85
- **MLP Ensemble**: 0.89
#### MSE AVM (Purple Bars)
- **XGBoost**: 0.35
- **LGBoost**: 0.60
- **MLP**: 1.00 (highest)
- **MLP Ensemble**: 0.85
#### MSE 1-day STD-DEV (Cyan Bars)
- **XGBoost**: 0.12
- **LGBoost**: 0.15
- **MLP**: 0.35
- **MLP Ensemble**: 0.20
### Key Observations
1. **Kendal Tau Correlation**: All models show strong correlation (>0.85), with XGBoost leading at 0.98.
2. **MSE AVM**: MLP has the highest error (1.00), while XGBoost performs best (0.35).
3. **MSE 1-day STD-DEV**: XGBoost has the lowest variability (0.12), followed by MLP Ensemble (0.20).
4. **MLP Ensemble**: Balances moderate Kendal Tau (0.89) with mid-range MSE metrics (0.85 AVM, 0.20 STD-DEV).
### Interpretation
- **Model Performance**: XGBoost excels in correlation and error metrics, suggesting robustness. MLP, while strong in correlation, shows high MSE AVM, indicating potential overfitting or instability.
- **MLP Ensemble**: Demonstrates a trade-off between correlation and error, performing better than individual MLP but worse than XGBoost/LGBoost in correlation.
- **Anomalies**: MLP’s high MSE AVM (1.00) contrasts with its moderate Kendal Tau (0.85), suggesting possible discrepancies in model stability or calibration.
- **Practical Implications**: XGBoost and LGBoost are optimal for high-correlation tasks, while MLP Ensemble may suit scenarios requiring balanced error metrics.
