Research overview: accelerometer behavioral classification in felids

This overview surveys how tri-axial acceleration data from collars are translated into behavioral states for cats, enabling activity budgets and the detection of key events like kills. Researchers typically collect a labeled training set, pairing observed or video-confirmed behaviors with their acceleration signatures, then train classifiers such as decision trees, random forests, or neural networks to assign states across the full record. Challenges include between-individual variation, the rarity of some behaviors, and the difficulty of obtaining ground-truth observations for secretive species. Validated classifiers reveal diel activity patterns, resting versus hunting time, and shifts in behavior around human disturbance. The overview frames behavioral classification as the analytical engine that turns raw acceleration into ecological meaning for felid movement studies.