Skip to main content

Table 1 Classifiers accuracy indexes for each composition and descriptor

From: Potential of machine learning and WorldView-2 images for recognizing endangered and invasive species in the Atlantic Rainforest

  

a

b

Precision

Recall

Fmeasure

Overall accuracy

Spectral attributes

ANN-RGB

CV

a=A. angustifolia

27

3

90.0%

96.4%

93.2%

Global accuracy

93.30%

b=H. dulcis

1

29

96.7%

90.6%

93.6%

Global error

6.70%

EV

a=A. angustifolia

14

1

93.3%

100.0%

96.7%

Global accuracy

96.70%

b=H. dulcis

0

15

100.0%

93.8%

96.9%

Global error

3.30%

MOD

a=A. angustifolia

29

1

96.7%

100.0%

98.3%

Global accuracy

98.30%

b=H. dulcis

0

30

100.0%

96.8%

98.4%

Global error

1.70%

RF-RBG

CV

a=A. angustifolia

28

2

93.3%

96.6%

94.9%

Global accuracy

95.00%

b=H. dulcis

1

29

96.7%

93.5%

95.1%

Global error

5.00%

EV

a=A. angustifolia

15

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

15

100.0%

100.0%

100.0%

Global error

0.00%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

ANN-NIR

CV

a=A. angustifolia

28

2

93.3%

93.3%

93.3%

Global accuracy

93.30%

b=H. dulcis

2

28

93.3%

93.3%

93.3%

Global error

6.70%

EV

a=A. angustifolia

14

1

93.3%

82.4%

87.8%

Global accuracy

86.70%

b=H. dulcis

3

12

80.0%

92.3%

86.2%

Global error

13.30%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

RF-NIR

CV

a=A. angustifolia

28

2

93.3%

96.6%

94.9%

Global accuracy

95.00%

b=H. dulcis

1

29

96.7%

93.5%

95.1%

Global error

5.00%

EV

a=A. angustifolia

15

0

100.0%

83.3%

91.7%

Global accuracy

90.00%

b=H. dulcis

3

12

80.0%

100.0%

90.0%

Global error

10.00%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

Textural attributes

ANN-RBG PHOG

CV

a=A. angustifolia

30

0

100.0%

56.6%

78.3%

Global accuracy

61.70%

b=H. dulcis

23

7

23.3%

100.0%

61.7%

Global error

38.30%

EV

a=A. angustifolia

14

1

93.3%

51.9%

72.6%

Global accuracy

53.30%

b=H. dulcis

13

2

13.3%

66.7%

40.0%

Global error

46.60%

MOD

a=A. angustifolia

30

0

100.0%

61.2%

80.6%

Global accuracy

68.30%

b=H. dulcis

19

11

36.7%

100.0%

68.3%

Global error

31.70%

RF-RBG PHOG

CV

a=A. angustifolia

30

0

100.0%

56.6%

78.3%

Global accuracy

61.70%

b=H. dulcis

23

7

23.3%

100.0%

61.7%

Global error

38.30%

EV

a=A. angustifolia

14

1

93.3%

51.9%

72.6%

Global accuracy

53.30%

b=H. dulcis

13

2

13.3%

66.7%

40.0%

Global error

46.70%

MOD

a=A. angustifolia

30

0

100.0%

61.2%

80.6%

Global accuracy

68.30%

b=H. dulcis

19

11

36.7%

100.0%

68.3%

Global error

31.70%

ANN-NIR PHOG

CV

a=A. angustifolia

30

0

100.0%

85.7%

92.9%

Global accuracy

91.70%

b=H. dulcis

5

25

83.3%

100.0%

91.7%

Global error

8.30%

EV

a=A. angustifolia

14

1

93.3%

70.0%

81.7%

Global accuracy

76.70%

b=H. dulcis

6

9

60.0%

90.0%

75.0%

Global error

23.30%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

RF-NIR PHOG

CV

a=A. angustifolia

29

1

96.7%

93.5%

95.1%

Global accuracy

95.00%

b=H. dulcis

2

28

93.3%

96.6%

94.9%

Global error

5.00%

EV

a=A. angustifolia

14

1

93.3%

87.5%

90.4%

Global accuracy

90.00%

b=H. dulcis

2

13

86.7%

92.9%

89.8%

Global error

10.00%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

ANN-RBG Edge

CV

a=A. angustifolia

24

6

80.0%

72.7%

76.4%

Global accuracy

75.00%

b=H. dulcis

9

21

70.0%

77.8%

73.9%

Global error

25.00%

EV

a=A. angustifolia

12

3

80.0%

75.0%

77.5%

Global accuracy

76.70%

b=H. dulcis

4

11

73.3%

78.6%

76.0%

Global error

23.30%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

RF-RBG Edge

CV

a=A. angustifolia

24

6

80.0%

80.0%

80.0%

Global accuracy

80.00%

b=H. dulcis

6

24

80.0%

80.0%

80.0%

Global error

20.00%

EV

a=A. angustifolia

14

1

93.3%

77.8%

85.6%

Global accuracy

83.30%

b=H. dulcis

4

11

73.3%

91.7%

82.5%

Global error

16.70%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

ANN-NIR Edge

CV

a=A. angustifolia

27

3

90.0%

75.0%

82.5%

Global accuracy

80.00%

b=H. dulcis

9

21

70.0%

87.5%

78.8%

Global error

20.00%

EV

a=A. angustifolia

10

5

66.7%

71.4%

69.0%

Global accuracy

70.00%

b=H. dulcis

4

11

73.3%

68.8%

71.0%

Global error

30.00%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

RF-NIR Edge

CV

a=A. angustifolia

25

5

83.3%

83.3%

83.3%

Global accuracy

83.30%

b=H. dulcis

5

25

83.3%

83.3%

83.3%

Global error

16.70%

EV

a=A. angustifolia

13

2

86.7%

81.3%

84.0%

Global accuracy

83.30%

b=H. dulcis

3

12

80.0%

85.7%

82.9%

Global error

16.70%

MOD

a=A. angustifolia

30

0

100.0%

100.0%

100.0%

Global accuracy

100.00%

b=H. dulcis

0

30

100.0%

100.0%

100.0%

Global error

0.00%

  1. Legend: CV, cross-validation; EV, external validation; and MOD using the supplied test set (all features)