Behavioural state dynamics
An animal can express so many behaviours. Is there some rhyme or rhythm to the way animals switch between behavioural states, and how long they stay in each state? Of course, it's not an easy question to answer -- it depends on the animals, and on the behaviours, in question. But these dynamics of behavioural states are a trait exhibited throughout the lifespan of an animal, and are likely very important from a fitness perspective. So, I worked with Ari Strandburg-Peshkin to find meaningful answers to this question using accelerometry data.
To do this, I constructed a behavioural classifier using a few Machine Learning algorithms. While behavioural classifiers have been built for tens of species over the past two and a half decades, there is no general purpose framework in deciding how to build a classifier if you want to. This is probably because of wide variation in data, arising in no small part because of the spectacular range of animals that have been tracked using such classifiers, all the way from meerkats to blue whales. I built several classifiers for the Spotted Hyena data Ari and her collaborators had collected, and finally obtained some long timescale data, using which I found bout-duration distributions, which I then summarised in my Master's thesis.
Now, I'm doing this with other species, and with a variety of behaviours. If there is some universal guideline to behavioural state dynamics, this work might help shed light on it.