Table 9 Results of the pairwise comparisons of seven models.

From: Example dependent cost sensitive learning based selective deep ensemble model for customer credit scoring

Metrics

Models

ECS-AdaBoost

ECSDNN

ECS-Stacking

CSNNE

CSCNN

CCS-CNN

AUC-ROC

ECS-SDE

0.01748

0.01381

0.04907

0.00693

0.55683

0.03423

ECS-AdaBoost

 

0.85011

0.70504

0.70504

0.09879

0.77972

ECSDNN

  

0.67454

0.77972

0.07003

0.70504

ECS-Stacking

   

0.49033

0.27420

0.85011

CSNNE

    

0.03423

0.55683

CSCNN

     

0.20760

AUC-PR

ECS-SDE

0.01748

0.00152

0.22850

0.02408

0.01748

0.22850

ECS-AdaBoost

 

0.42581

0.22850

0.93959

1.00000

0.22850

ECSDNN

  

0.04907

0.35957

0.42581

0.04907

ECS-Stacking

   

0.27882

0.22850

1.00000

CSNNE

    

0.93959

0.27882

CSCNN

     

0.22850

BS+

ECS-SDE

0.01748

0.21092

0.01381

0.01381

0.59022

0.27882

ECS-AdaBoost

 

0.21092

0.85011

0.82303

0.08170

0.16986

ECSDNN

  

0.16986

0.13716

0.59022

0.85011

ECS-Stacking

   

0.85011

0.05888

0.13716

CSNNE

    

0.04279

0.11291

CSCNN

     

0.70504

Save

ECS-SDE

0.01381

0.17635

0.15807

0.03210

0.15807

0.00330

ECS-AdaBoost

 

0.39034

0.44947

0.82303

0.44947

0.70504

ECSDNN

  

0.89261

0.51706

0.89261

0.17635

ECS-Stacking

   

0.59022

1.00000

0.20760

CSNNE

    

0.59022

0.51706

CSCNN

     

0.20760

BS

ECS-SDE

0.59022

0.08782

0.59022

0.74073

0.02853

0.06127

ECS-AdaBoost

 

0.02408

0.22850

0.74073

0.01406

0.01748

ECSDNN

  

0.30026

0.04206

0.70504

0.85011

ECS-Stacking

   

0.38526

0.12344

0.22850

CSNNE

    

0.01748

0.02853

CSCNN

     

0.74073

  1. Bold values indicate that the adjusted p-value is less than 0.05.