Phenomics

Modeling Behavioral Patterns in Group-Housed Pigs Using Continuous Trajectories

---

Abstract

With global demand for animal products expected to rise by 70% by 2050, sustainable and humane livestock production is more urgent than ever. In this context, behavioral phenotypes are emerging as promising indicators of animal health, welfare, and adaptability. Using individual movement patterns, it is possible to derive objective behavioral measures. In this work, we introduce the use of Hidden Markov Models (HMMs) to infer behavioral states from animal trajectories of group-housed nursery pigs. Our objective was to identify distinct behavioral states and quantify individual variation in state-switching dynamics to study behavioral patterns.Animal's x and y positions were tracked every 1s for 10 individually identified pigs over two consecutive hours starting at 6:25 AM of the second day post-mixing. Pig positions were obtained from overhead cameras recording animal movement. The center of the body mass was extracted from images and used to track individuals. Pig positions were converted into movement variables, specifically, step length and turning angle. Hidden Markov models were fitted with varying numbers of states, and model selection was guided by both Akaike Information Criterion (AIC) and biological interpretability. The best-fitting hidden Markov models included 4 movement states shared across animals but with individual transition probabilities. The estimated movement parameters (Table 1) characterized each movement state. Comparing transition probabilities across all 10 pigs, the probability of remaining in the same state was highest for State 1 (mean ± SD = 0.92 ± 0.06), followed by State 3 (0.80 ± 0.01), State 2 (0.75 ± 0.03), and State 4 (0.74 ± 0.06). The most extreme values included individuals with the highest probability of remaining in State 1 (96.0%) and the lowest probability of remaining in State 4 (68.3%), illustrating individual differences in movement transition traits. On average, pigs spent 0.34 ± 0.08 of their time in State 2, 0.34 ± 0.05 in State 3, 0.18 ± 0.10 in State 1, and 0.13 ± 0.07 in State 4. Variation among individuals was evident, with some pigs allocating the highest proportions of time to State 2 (51-29%), whereas others spent the highest proportions of time in State 3 (40-33%) or State 1 (33%).The estimated variance for transitions indicated no variation (SD=0) in transitioning from state 1 to state 4, while the highest variance of 0.07 was observed in transitioning from state 1 to state 3 across individuals. Overall, the HMM identified four movement states, with pigs allocating different amounts of time to each state. While most pigs showed similar transition patterns, some transitions showed individual variability, indicating differences in individual-level locomotion patterns.

Keywords: 2026

How to Cite:

Shumaly, M., Bi, Y., Acharya, B., Rosero, D., Johnson, A. & Steibel, J., (2026) “Modeling Behavioral Patterns in Group-Housed Pigs Using Continuous Trajectories”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2285980. doi: https://doi.org/10.31274/wcgalp.23829

Rights: 1