Skip to main content

Table 5 RRS model results for the best twenty models using feature selection and best ten all variable (AV) models. Feature subset numbers are defined in Table 3. For AV, feature set indicates the sensor and number of variables (in parentheses) in the subset. Results are mean ± standard deviation

From: Feature selection for elderly faller classification based on wearable sensors

Method

Feature Set

Modela

Accuracy (%)

Sensitivity (%)

Specificity (%)

PPV (%)

NPV (%)

F1

MCC

SR

CFS/FCBF

9

SVM-2

77.9 ± 4.8

26.4 ± 15.9

95.1 ± 5.2

64.6 ± 32.6

79.7 ± 3.5

0.355 ± 0.182

0.305 ± 0.202

55

Relief-F

1

SVM-7

74.0 ± 8.1

44.3 ± 20.2

83.3 ± 9.1

47.3 ± 20.1

82.8 ± 5.5

0.441 ± 0.173

0.286 ± 0.218

58

CFS/FCBF

9

SVM-3

78.0 ± 4.9

25.5 ± 15.6

95.5 ± 5.1

65.4 ± 33.5

79.5 ± 3.5

0.348 ± 0.183

0.304 ± 0.205

59

Relief-F

3

NN-21

75.3 ± 6.9

32.0 ± 20.4

89.7 ± 8.6

50.6 ± 29.7

80.1 ± 4.8

0.367 ± 0.203

0.259 ± 0.225

61

CFS/FCBF

9

NN-8

76.9 ± 6.0

27.6 ± 18.1

93.4 ± 8.0

60.2 ± 34.5

79.7 ± 4.0

0.349 ± 0.193

0.287 ± 0.212

62

Relief-F

1

SVM-5

73.6 ± 8.0

45.3 ± 20.1

82.6 ± 9.0

46.6 ± 19.1

82.9 ± 5.5

0.443 ± 0.169

0.285 ± 0.214

62

CFS/FCBF

9

NN-10

76.7 ± 6.0

28.2 ± 18.6

92.9 ± 7.9

58.3 ± 33.8

79.7 ± 4.1

0.351 ± 0.197

0.282 ± 0.214

64

Relief-F

5

SVM-4

74.6 ± 7.5

37.5 ± 20.2

86.3 ± 8.5

47.8 ± 23.9

81.6 ± 5.1

0.401 ± 0.188

0.262 ± 0.226

64

Relief-F

1

NN-21

76.0 ± 6.9

31.5 ± 20.2

90.1 ± 8.4

49.7 ± 30.1

80.9 ± 4.6

0.362 ± 0.205

0.258 ± 0.227

66

Relief-F

3

NN-25

75.3 ± 6.8

31.9 ± 20.6

89.7 ± 8.6

50.7 ± 29.5

80.1 ± 4.9

0.365 ± 0.202

0.257 ± 0.223

67

Relief-F

1

SVM-6

74.0 ± 7.6

38.5 ± 20.1

85.2 ± 8.7

46.5 ± 22.2

81.7 ± 5.2

0.402 ± 0.180

0.255 ± 0.219

71

Relief-F

3

NN-23

75.2 ± 6.8

31.9 ± 20.5

89.6 ± 8.5

50.3 ± 29.3

80.1 ± 4.8

0.365 ± 0.202

0.255 ± 0.225

74

Relief-F

2

NN-15

75.2 ± 7.1

30.3 ± 19.7

90.2 ± 8.8

50.9 ± 31.1

79.7 ± 4.7

0.356 ± 0.204

0.252 ± 0.230

76

Relief-F

8

NB-Q

68.3 ± 8.9

55.7 ± 20.6

72.5 ± 11.9

41.5 ± 13.6

83.5 ± 6.6

0.461 ± 0.140

0.264 ± 0.192

84

CFS/FCBF

9

NB-Q

70.9 ± 7.9

41.3 ± 22.4

80.7 ± 9.8

41.5 ± 20.0

80.9 ± 6.0

0.397 ± 0.184

0.221 ± 0.222

100

Relief-F

3

SVM-3

70.9 ± 8.0

37.9 ± 20.0

81.9 ± 9.6

42.2 ± 20.5

80.1 ± 5.5

0.381 ± 0.173

0.208 ± 0.214

102

AV

I(30),H(29), P(29)

SVM-3

75.5 ± 5.6

21.2 ± 16.2

93.6 ± 5.7

49.9 ± 35.2

78.2 ± 3.7

0.282 ± 0.195

0.207 ± 0.225

108

Relief-F

4

NN-9

73.4 ± 7.2

27.7 ± 19.5

88.7 ± 9.1

44.0 ± 29.8

78.8 ± 4.6

0.318 ± 0.201

0.196 ± 0.226

119

Relief-F

6

SVM-4

71.9 ± 7.3

31.7 ± 18.7

85.3 ± 8.2

42.2 ± 23.3

79.1 ± 4.8

0.346 ± 0.180

0.188 ± 0.218

120

Relief-F

4

NN-21

73.7 ± 6.8

25.3 ± 19.2

89.8 ± 8.6

43.0 ± 31.2

78.5 ± 4.4

0.298 ± 0.203

0.185 ± 0.225

126

Relief-F

7

NN-21

72.9 ± 6.9

25.8 ± 18.8

88.6 ± 9.0

41.7 ± 29.4

78.3 ± 4.4

0.298 ± 0.194

0.173 ± 0.218

139

AV

I(30),P(29), LS(29)

SVM-2

70.2 ± 7.1

30.8 ± 18.5

83.3 ± 8.7

38.4 ± 21.4

78.5 ± 4.7

0.326 ± 0.170

0.153 ± 0.204

139

AV

I(30),H(29)

SVM-4

74.2 ± 5.0

12.3 ± 13.2

93.7 ± 5.2

33.1 ± 35.8

77.2 ± 2.9

0.170 ± 0.175

0.087 ± 0.214

151

AV

H(29)

SVM-4

73.3 ± 5.8

16.1 ± 14.6

91.4 ± 6.4

35.1 ± 32.3

77.6 ± 3.3

0.209 ± 0.178

0.101 ± 0.215

153

AV

I(30),P(29)

SVM-2

67.6 ± 7.2

32.1 ± 18.4

79.4 ± 9.0

33.9 ± 17.6

78.0 ± 4.8

0.318 ± 0.159

0.117 ± 0.193

154

AV

I(30),P(29), LS(29),RS(29)

NB-Q

60.8 ± 9.2

37.6 ± 20.5

68.6 ± 11.7

28.3 ± 14.1

76.9 ± 6.4

0.314 ± 0.152

0.057 ± 0.203

171

AV

I(30),P(29)

NN-9

68.0 ± 8.3

24.4 ± 18.6

82.5 ± 10.6

31.4 ± 24.2

76.7 ± 4.8

0.258 ± 0.179

0.077 ± 0.216

178

AV

H(29)

SVM-2

67.8 ± 7.2

24.7 ± 17.3

81.4 ± 8.8

29.1 ± 19.7

77.5 ± 4.3

0.256 ± 0.163

0.063 ± 0.196

181

AV

I(30),H(29), P(29),LS(29)

NN-20

67.2 ± 7.9

21.0 ± 16.9

82.6 ± 10.4

27.8 ± 22.7

75.9 ± 4.3

0.226 ± 0.167

0.041 ± 0.199

190

AV

H(29),P(29), LS(29),RS(29)

NN-5

65.3 ± 8.2

16.5 ± 15.2

81.5 ± 11.0

22.3 ± 22.2

74.5 ± 4.0

0.177 ± 0.153

−0.021 ± 0.190

201

  1. AV all variables, I pressure-sensing insole measures, H head accelerometer measures, P pelvis accelerometer measures, LS left shank accelerometer measures, RS right shank accelerometer measures, NN neural network, NB naïve Bayesian model, SVM support vector machine, SR summed rank
  2. aNN-a, where a is the number of nodes in the hidden layer; SVM-b, where b is the polynomial degree; NB-Q is quadratic naïve Bayesian