Skip to main content

Table 6 Results of the models on the balanced data

From: Machine learning with asymmetric abstention for biomedical decision-making

Name

Abstention

MCC

Rejected

Interval

wdbc

None

0.93

–

–

wdbc

Symmetric

0.952

< 5%

[0.43, 0.57]

wdbc

Asymmetric

0.952

< 5%

[0.427, 0.467]

pc

None

0.438

–

–

pc

Symmetric

0.558

< 10%

[0.44, 0.56]

pc

Asymmetric

0.558

< 10%

[0.439, 0.539]

ctg

None

0.955

–

–

ctg

Symmetric

0.977

< 5%

[0.36, 0.64]

ctg

Asymmetric

0.955

< 1%

[0.354, 0.414]

  1. Abstention represents the abstention approach, i.e., none, symmetric, or asymmetric abstention. Rejected refers to the percentage of rejected samples, and interval represents the bounderies of the abstention interval