Table 2 Mean variable importance by modelling method, as a percentage.

From: Predicting global invasion risks: a management tool to prevent future introductions

Model

Bio1

Bio2

Bio3

Bio8

Bio14

Bio18

Bio19

TpH

INMSR

TWIS

ANN

12.50

4.33

8.96

6.79

30.11

12.43

16.46

0.42

2.54

5.44

GLM

3.05

2.89

22.36

6.63

41.50

1.17

5.00

0.79

4.37

12.24

GBM

4.36

2.42

25.89

1.65

46.73

1.63

1.08

0.03

2.05

14.16

SRE

10.08

10.12

10.11

10.45

7.28

9.60

9.37

10.52

11.25

11.22

CTA

9.87

5.64

20.10

4.67

34.33

5.24

3.60

0.60

3.52

12.43

RF

12.62

5.62

24.48

3.98

18.69

6.21

6.15

2.51

7.86

11.88

MARS

7.50

5.48

17.18

5.25

41.93

1.26

6.96

0.10

3.10

11.24

FDA

7.48

3.50

22.83

5.42

35.04

1.42

4.70

0.38

6.70

12.54

GAM

7.85

4.01

30.64

6.74

15.49

3.75

11.60

1.27

5.61

13.05

EMmw

8.15

4.23

21.60

5.13

32.94

4.13

6.93

0.77

4.48

11.64

  1. Variable importance scores, as measured by randomization technique, calculated for individual models as 1— Pearson’s correlation between predictions, before and after randomization. Scores were then converted into a % of the sum of all variable importance scores for each modelling method. The three most important variables in the final EMmw model are highlighted in bold.
Back to article page