Research overview: machine learning for felid movement classification

As tracking datasets have grown to millions of points and multi-sensor streams, machine learning has become increasingly central to interpreting them. This overview surveys the application of supervised classifiers, random forests, support vector machines, and deep neural networks, to label behaviors from accelerometer and movement data, as well as unsupervised methods to discover behavioral clusters. For cats, machine learning detects kills, classifies activity states, and flags anomalous movement that may indicate disturbance or mortality. The overview discusses the need for labeled training data, the risk of overfitting to particular individuals or sites, and the importance of interpretability for ecological inference. By scaling pattern recognition to enormous biologging datasets, machine learning extends what researchers can extract from felid telemetry, while raising new questions about validation and transferability.