Research shows how machine learning can discern social grooming behavior from accelerometer signals in wild baboons

Scientists at Swansea University and the University of Cape Town used collar-mounted accelerometers to track the social grooming behavior of wild baboons.

Research published in journals Royal Society Open Sciencewas the first investigator to successfully calculate a grooming budget using this method, paving the way for a whole range of future research directions.

Using accelerometer-equipped collars made at Swansea University, the team recorded the activity of baboons in Cape Town, South Africa, and recorded common behaviors such as resting, walking, foraging, running and grooming. activities were identified and quantified.

A supervised machine learning algorithm was trained on matching acceleration data from video recordings of baboons and was successful in recognizing grooming giving and receiving with high overall accuracy.

The team then applied machine learning models to acceleration data collected from 12 baboons to continuously quantify grooming and other behaviors throughout the day and night.

Lead author Dr Charlotte Christensen of the University of Zurich said: Social behavior of animals, especially non-human primates. ”

Social grooming is one of the most important social behaviors of primates and has been central to primatological research since the 1950s.

Previously, scientists relied on direct observation to determine how well primates groomed. Direct observation provides systematic data, but it is sparse and non-sequential, with the additional limitation that researchers can only see a few animals at a time. .

Techniques like the one used in this study are revolutionizing the field of animal behavior research and enabling exciting new areas of research.

Senior author Dr Ines Fürtbauer of Swansea University said: It is about the formation and maintenance of social bonds and the mechanisms that underpin the relationship between sociality and health and fitness. ”