Skip to main content

Table 4 F-measure one-versus-all classification

From: Towards a supervised classification of neocortical interneuron morphologies

 

Cell Type

Classifier

FSS

Sampling

F-measure

TPR

TNR

Morphom.

1

redChC

RMLR

  

0.40

2 / 7

209 / 210

11

2

 

RF

 

Yes

0.49

2.8 / 7

208.5 / 210

103

3

 

GBM

KW

 

0.50

3 / 7

208 / 210

15

4

 

RF

KW

Yes

0.46

3.5 / 7

205.2 / 210

15

5

redBTC

NB

  

0.35

8 / 15

179 / 202

103

6

 

GBM

 

Yes

0.36

5.8 / 15

191.2 / 202

23

7

 

LDA

KW

 

0.44

6 / 15

196 / 202

7

8

 

LDA

KW

Yes

0.40

8.8 / 15

181.8 / 202

7

9

orangeDBC

RMLR

  

0.70

15 / 22

189 / 195

17

10

 

RF

 

Yes

0.70

14.6 / 22

189.8 / 195

103

11

 

RF

RF BVI

 

0.72

13 / 22

194 / 195

6

12

 

RF

KW

Yes

0.70

15.4 / 22

188.2 / 195

61

13

orangeSBC

CART

  

0.63

16 / 28

182 / 189

5

14

 

RF

 

Yes

0.66

20.6 / 28

174.8 / 189

103

15

 

NNET

RF BVI

 

0.74

21 / 28

181 / 189

7

16

 

RF

RF BVI

Yes

0.69

22.5 / 28

173.8 / 189

7

17

greenNBC

CART

  

0.73

32 / 44

161 / 173

4

18

 

RF

 

Yes

0.81

36.2 / 44

164 / 173

103

19

 

GBM

RF BVI

 

0.78

34 / 44

164 / 173

9

20

 

GBM

RF BVI

Yes

0.77

37.4 / 44

157 / 173

9

21

greenMC

SVM

  

0.77

37 / 50

158 / 167

103

22

 

RF

 

Yes

0.81

40.2 / 50

158.4 / 167

103

23

 

RMLR

KW

 

0.80

38 / 50

160 / 167

22

24

 

RF

KW

Yes

0.82

40.9 / 50

157.8 / 167

62

25

orangeLBC

GBM

  

0.61

26 / 51

158 / 166

103

26

 

RF

 

Yes

0.67

29.8 / 51

157.4 / 166

103

27

 

GBM

RF BVI

 

0.66

31 / 51

154 / 166

4

28

 

GBM

RF BVI

Yes

0.67

37.4 / 51

142.2 / 166

4

29

greenBA

RF

  

0.86

106 / 123

76 / 94

103

30

 

SVM

 

Yes

0.86

101.9 / 123

80.8 / 94

103

31

 

SVM

KW

 

0.88

105 / 123

84 / 94

68

32

 

SVM

KW

Yes

0.88

104.2 / 123

84.2 / 94

68

  1. The table shows, for each type, the best F-measure in all four learning settings: with and without sampling, and with and without feature selection. TPR: true positive rate; TNR: true negative rate; the minority class is always the positive one, except for BA; Morphom.: the number of morphometrics in the model. Types are sorted from least to most frequent (e.g., ChC, with only seven examples, is shown uppermost). The best F-measure for each type is typeset in bold. Types with their best F-measure ≥0.75 are shown in green; those with an F-measure ≥0.60 in orange; and the rest in red