Skip to main content

Table 3 The mean precision, recall and F1 statistic (± SD) for each random forest model predicting behaviour or behaviour including slope of terrain for Alpine ibex and pygmy goats

From: Limitations of using surrogates for behaviour classification of accelerometer data: refining methods using random forest models in Caprids

Model

Classification accuracy

Mean precision

Mean recall

Mean F1 statistic

Random split behaviour

98.3%

0.95 ± 0.05

0.98 ± 0.018

0.96 ± 0.030

Random split behaviour including slope of terrain

98.2%

0.95 ± 0.042

0.98 ± 0.018

0.96 ± 0.024

Data split per individual behaviour

63.0%

0.48 ± 0.32

0.55 ± 0.27

0.46 ± 0.28

Data split per individual behaviour including slope of terrain

68.1%

0.36 ± 0.028

0.42 ± 0.082

0.034 ± 0.046

Pygmy goat predicting Alpine ibex behaviour

55.5%

0.55 ± 0.39

0.62 ± 0.10

0.48 ± 0.30

Pygmy goat predicting female Alpine ibex behaviour

60.2%

0.55 ± 0.39

0.50 ± 0.26

0.49 ± 0.32

Pygmy goat predicting Alpine ibex behaviour including slope of terrain

59.8%

0.29 ± 0.38

0.30 ± 0.29

0.27 ± 0.32

Pygmy goat predicting female Alpine ibex behaviour including slope of terrain

67.8%

0.28 ± 0.42

0.25 ± 0.34

0.26 ± 0.37

Randomly generated behaviours

15.4%

0.010 ± 0.27

0.058 ± 0.09

0.038 ± 0.08

Randomly generated behaviours with slope of terrain

26.4%

0.068 ± 0.24

0.040 ± 0.11

0.041 ± 0.12