feat: Add ML-based Risk Scoring to Risk Agent

[niksacdev]

Description

Enhance the Risk Agent with machine learning capabilities for more sophisticated risk assessment and decision making.

Acceptance Criteria

• Integrate ML model for risk scoring
• Train model on historical loan performance data
• Implement feature engineering pipeline
• Add explainable AI for decision transparency
• Create model monitoring and retraining pipeline
• Maintain fallback to rule-based system

Technical Details

• Use gradient boosting (XGBoost/LightGBM) for risk modeling
• Feature engineering from all agent assessments
• SHAP values for explainability
• A/B testing framework for model comparison
• Model versioning and rollback capability

Expected Outcome

• Improve decision accuracy by 15-20%
• Reduce false positives in risk detection
• Provide clear explanations for decisions

Priority

Low - Future enhancement after core system stable