Skip to main content

Table 6 Scorecard phase 2: computing the supervisory riskiness for regulatory capital

From: Measuring the model risk-adjusted performance of machine learning algorithms in credit default prediction

  Lasso CART RF XGBoost MLP
Statistics 1 3 3 2 4
 Capital: 7.9%      
 Credit: 6.41%      
 Provisions: 6.38%      
Technology 1 1 3 2 5
 Capital: 2.5%      
 Credit: 3.09%      
 Provisions: 2.36%      
Market conduct 1 1 2 2 3
 Capital: 3.75%      
 Credit: 5.13%      
 Provisions: 3.9%      
Supervisory model risk capital 14.15 29.95 38.7 28.3 55.35
Supervisory model risk credit 14.63 27.45 38.76 29.26 56.48
Supervisory model risk provisions 12.64 25.4 34.02 25.28 49,02