Behaviour Classification on Giraffes (Giraffa camelopardalis) Using Machine Learning Algorithms on Triaxial...

Area of Science:

• Wildlife ecology
• Conservation technology
• Animal behavior analysis

Background:

• Giraffes (Giraffa camelopardalis) are vulnerable, impacting African ecosystems.
• Conservation requires monitoring wildlife behavior.
• Accelerometers offer remote wildlife tracking capabilities.

Purpose of the Study:

• To evaluate the accuracy of two accelerometers (e-obs and AWT) for classifying giraffe behavior.
• To assess the suitability of these devices for conservation research.

Main Methods:

• Two high-resolution accelerometers (e-obs, AWT) were attached to captive giraffes.
• Body movement data was collected and analyzed using the Random Forests algorithm.
• Accuracy of automatic behavior classification was compared for different behaviors.

Main Results:

• Both accelerometers achieved high prediction accuracy for behaviors with limited head/neck movement (feeding, drinking).
• Accuracy was lower for behaviors involving diverse postures (standing, rumination).
• The AWT device requires further adaptation for wild applications.

Conclusions:

• Accelerometers show potential for giraffe behavioral classification and conservation.
• Combined with GPS, these devices can significantly aid giraffe conservation strategies.
• Further technological refinement is needed for optimal field application.