Unsupervised exploratory cluster analysis of free range laying hens to determine the use of aviary feed chains and range access

Abstract

Clustering is a common data mining methodology used for improved subject understanding. The aim of this study was to identify sub-populations oflaying hens housed in an aviary system to understand the use of feeding chains which can affect hen performance and welfare. A total of 5,641 Lohmann Brown free-range laying hens placed amongst 3 commercial flocks equipped with a 3-tier aviary system were individually monitored using radio- frequency identification (RFID) technology. Individual body weights of all hens were obtained at 16, 22, and 72 weeks of age. K-Means cluster analysis optimised with the Calinski-Harabasz Criterion was performed. Hens of cluster 1 (n=2442 hens) spent significantly more time on the lower tier feeding chain (14.5 ± 2.36 hours/hen/day) compared to hens of cluster 2 (n=2083; 6.9 ±2.4 h/hen/day) and hens of cluster 3 (n=1116; 2.0 ± 1.9 h/hen/day), respectively (P < 0.05). Hens of cluster 3 spent 10.9 ± 3.6 h/hen/day at the top tier feeder chain compared to hens of cluster 1 and 2 (0.9 ± 1.1 and 3.6 ±2.1 h/hen/day respectively; P < 0.05). Hens of all clusters were of comparable body weight distributions at week 16, 22 and 72 weeks of age. Hens of cluster 3 spent the least time on the range and the most time on the upper tier feed chain of the upper tier (P < 0.05), however there was no significant impact on weight gain between 16 and 72 weeks. We conclude that several subpopulations of hens can be identified in the aviary system and that these subpopulations result in an uneven load on the resources (e.g feed chains). Further analysis of the data using classification models based on support vector machines, artificial neural networks and decision trees is warranted to predict other parameter of hen performance.