Associating panting levels of broiler breeders in a commercial battery cage barn with environmental and age factors via machine learning modeling.

Panting may serve as one of the most distinct behavioral indicators of heat stress for poultry without sweat glands. Broiler breeders raised in battery cages suffer from heat stress during hot weather seasons, because they have limited movement to seek cooler zones inside cages, and high cage stocking densities can prevent heat dissipation. Easily accessible but impactful environmental and age factors could be used to associate with panting situations, which provides insights into heat stress management. The objective of this study was to associate panting situations of broiler breeders in a commercial battery cage barn with environmental and age factors. Data were collected when the broiler breeder hens were 103-147 days of age. The monitored intra-cage environmental parameters included wind speed (0.09-0.57 m/s), temperature (23.03-34.72 °C), and relative humidity (55.73%-89.83%), alongside the corresponding panting ratio (0%-22%). The commercial barn contained 15,200 Arbor Acres broiler breeder hens, and the data collection was conducted during summer in the middle part of China. A total of 1,098 data points were collected. Two sets of machine learning modeling were operated, which were regressing continuous panting ratio and classifying three levels (low, medium, and high based on unsupervised K-means clustering) of panting. A total of 13 machine learning algorithms were comparatively examined with 80% data for training and 20% for testing. Number of features was expanded from 4 to 13, and then the expanded features with the Pearson correlation coefficient less than 0.1 were removed. The best regression performance was 0.286 R