Skip to main content

Table 4 LOIO cross-validation results

From: Behavioural compass: animal behaviour recognition using magnetometers

Sensor

Vigilance

Resting

Foraging

Running

Overall Accuracy (%)

Sen.

(%)

Spec.

(%)

Prec.

(%)

Sen.

(%)

Spec.

(%)

Prec.

(%)

Sen.

(%)

Spec.

(%)

Prec.

(%)

Sen.

(%)

Spec.

(%)

Prec.

(%)

Magnetometer

95.2

± 2.4

97.9

± 1.9

95.2

± 6.2

65.4

± 25.9

98.9

± 0.9

77.3

± 31.1

98.4

± 0.9

97.0

± 1.2

95.5

± 0.5

86.5

± 3.7

100

± 0.0

96.4

± 3.4

96.0

± 1.5

Accelerometer

95.8

± 2.8

98.4

± 1.2

96.4

± 4.5

71.4

± 23.6

98.9

± 1.2

81.1

± 28.0

98.8

± 1.0

97.4

± 1.5

95.3

± 7.0

86.3

± 13.2

99.9

± 0.1

89.1

± 11.1

96.5

± 1.8

  1. The performance of the SVM-SVM-SVM hybrid model with magnetometer data is benchmarked against that obtained with accelerometer data reported in [10]. Performance metrics were calculated separately for each test individual, and their mean and standard deviation across test individuals are shown here. SVM: Support Vector Machine