Skip to main content

Table 7 Varying levels of class imbalance with ROS and RUS

From: Medicare fraud detection using neural networks

Method

\(n_{neg}\)

\(n_{pos}\)

\(N_{train}\)

\(n_{neg}{:}n_{pos}\)

–

3,377,421

1085

3,378,506

99.97:0.03

RUS-1

107,402

1085

108,487

99:1

RUS-2

4390

1085

5475

80:20

RUS-3

1620

1085

2705

60:40

RUS-4

1085

1085

2170

50:50

RUS-5

710

1085

1795

40:60

ROS-1

3,377,421

33,635

3,411,046

99:1

ROS-2

3,377,421

844,130

4,221,551

80:20

ROS-3

3,377,421

2,251,375

5,628,796

60:40

ROS-4

3,377,421

3,377,421

6,754,842

50:50

ROS-5

3,377,421

5,064,780

8,442,201

40:60

ROS–RUS-1

1,688,710

1,688,710

3,377,420

50:50

ROS–RUS-2

844,355

844,355

1,688,710

50:50

ROS–RUS-3

337,742

337,742

675,484

50:50