Keywords: deep learning, random forests, animal movement, recurrent, time series
Animal movement drives important ecological processes such as migration and the spread of infectious disease. Current approaches to modeling animal tracking data focus on parametric models used to understand environmental effects on movement behavior and to fill in missing tracking data. Machine Learning and Deep learning algorithms are powerful and flexible predictive modeling tools but have rarely been applied to animal movement data. In this study we compare several architectures of random forests and neural networks for modeling and predicting animal movement. Our approach considers predicting an animal's velocity at each time step using the position and velocity of the animal at several previous time lags as well as other derived variables from the current state of the system. Stationary behavior is captured by modeling each animal's decision to move or remain in place at each time point as a categorical variable which is also predicted using these same time-lagged covariates. We also explore the use of recurrent neural networks to capture long term repetitive behavior of the animals. We apply this approach to high resolution ant movement data and obtain stochastic simulations of ant colonies.