Skip to main content

Table 5 The effectiveness of 17 different machine learning methods in the testing set

From: Research on imbalance machine learning methods for MR\(T_1\)WI soft tissue sarcoma data

N

FS

ST

CM

AUC ± \({\sigma }\)

Acc(%) ± \({\sigma }\)

Sens(%) ± \({\sigma }\)

Spec(%) ± \({\sigma }\)

G-mean ± \({\sigma }\)

1

RFE

ROSE

ERT

0.6013 ± 0.0482

78.82 ± 0.0545

98.66 ± 0.0227

21.60 ± 0.1014

0.4477 ± 0.1154

2

RFE

SMOTE

ERT

0.6863 ± 0.0515

81.37 ± 0.0500

95.80 ± 0.0284

41.47 ± 0.0972

0.6260 ± 0.0782

3

RFE

STT

ERT

0.6879 ± 0.0553

81.57 ± 0.0533

96.03 ± 0.0254

41.55 ± 0.1091

0.6263 ± 0.0860

4

RFE

ADASYN

ERT

0.6461 ± 0.0595

79.41 ± 0.0464

95.04 ± 0.0279

34.18 ± 0.1121

0.5621 ± 0.1017

5

RFE

ROSE

RF

0.6197 ± 0.0473

77.45 ± 0.0533

93.97 ± 0.0425

29.97 ± 0.0865

0.5258 ± 0.0746

6

RFE

SMOTE

RF

0.6567 ± 0.0488

76.27 ± 0.0502

87.50 ± 0.0427

43.84 ± 0.1032

0.6147 ± 0.0700

7

RFE

STT

RF

0.6580 ± 0.0447

76.67 ± 0.0448

88.35 ± 0.0396

43.25 ± 0.1018

0.6133 ± 0.0680

8

RFE

ADASYN

RF

0.6142 ± 0.0618

73.92 ± 0.0599

87.60 ± 0.0482

35.24 ± 0.1026

0.5503 ± 0.0877

9

RFE

ROSE

BRF

0.6151 ± 0.0332

77.45 ± 0.0446

94.52 ± 0.0356

28.49 ± 0.0645

0.5154 ± 0.0593

10

RFE

SMOTE

BRF

0.6287 ± 0.0487

74.90 ± 0.0422

86.97 ± 0.0381

38.77 ± 0.1031

0.5750 ± 0.0770

11

RFE

STT

BRF

0.6367 ± 0.0578

75.69 ± 0.0436

87.84 ± 0.0461

39.51 ± 0.1182

0.5822 ± 0.0872

12

RFE

ADASYN

BRF

0.6243 ± 0.0331

74.12 ± 0.0441

86.76 ± 0.0370

38.10 ± 0.0735

0.5720 ± 0.0503

13

RFE

ROSE

SVM

0.6863 ± 0.2226

77.65 ± 0.0436

87.49 ± 0.0438

52.30 ± 0.1295

0.6715 ± 0.0789

14

RFE

SMOTE

SVM

0.6812 ± 0.0591

76.47 ± 0.0606

85.41 ± 0.0633

50.82 ± 0.0894

0.6564 ± 0.0672

15

RFE

STT

SVM

0.6812 ± 0.0591

76.47 ± 0.0606

85.41 ± 0.0633

50.82 ± 0.0894

0.6564 ± 0.0672

16

RFE

ADASYN

SVM

0.6795 ± 0.0483

75.29 ± 0.0499

83.48 ± 0.0672

52.43 ± 0.0815

0.6588 ± 0.0557

17

RFE

GDO

SVM

0.6691 ± 0.0685

76.67 ± 0.0657

87.51 ± 0.0557

46.30 ± 0.1083

0.6328 ± 0.0580

  1. Best results are highlighted in bold style
  2. N number; FS feature selection; ST sampling technique; CM classification method; AUC area under the curve; Sens sensitivity; Spec specificity; ROSE random oversampling examples; SMOTE synthetic minority oversampling technique; STT SMOTETomek; ADASYN adaptive synthetic sampling; RFE recursive feature elimination; ERT extremely randomized trees; RF random forest; BRF balanced random forest; SVM support vector machine