Skip to main content

Table 16 Average cost-sensitive results (30 runs)

From: Medicare fraud detection using neural networks

Method

\(w_{pos}\)

\(w_{neg}\)

Hidden layers

Decision threshold

ROC AUC

TPR

TNR

G-Mean

CoSen-1-2

3112

1

2

0.4760

0.8075

0.7574

0.7031

0.7283

CoSen-1-4

  

4

0.4480

0.8011

0.8008

0.6522

0.7213

CoSen-2-2

0.9997

0.0003

2

0.4780

0.8072

0.7529

0.7086

0.7291

CoSen-2-4

  

4

0.4825

0.8028

0.7265

0.7389

0.7318

  1. Italic font indicates the maximum ROC AUC score