Understanding animal behavior using their trajectories: A case study of gender specific trajectory trends

Abstract

Generally, behavior and movement could be closely attributed with each other. In other words, in most cases, behavior expressed as a set of specific movement patterns in time. These movement traces or trajectories representing behavior could provide a window into the underlying state of the subjects. In this study, analogies drawn between text and trajectories which allowed us to employ sentiment analysis and topic model methods to analyze trajectories. It is assumed that trajectories consist of key points which are commonly and frequently traversed. It is proposed that analogously in trajectory analysis, key points frequency would encapsulate information about the subject or the key points in trajectory generated by latent distribution which attributed to certain behavior or specific group of subjects with similar behavioral features. To test this hypothesis, an experiment was conducted which examines the influence of gender in composition of key points in birds’ trajectories logged from a seabird species called Streaked Shearwater Calonectris leucomelas. It was shown that genders have specific distribution over the key points. Therefore, key points membership in trajectory could be attributed to a specific gender and even a simple classifier would provide information about the gender of the subject simply by observing the trajectory’s key points. It was concluded that like text, trajectories composed of smaller elements which could be associated to a specific latent state. Learning or exploiting these associations revealed essential information about identity and behavior of the subject of observation.