In commercial swine farming, the ability to predict the productivity of sows is extremely valuable as it provides herd management guidance, improving the farm cost efficiency. However, sow productivity is a multifaceted phenotype that involves complex patterns between various measurable sow attributes. The objective of this study was to compare and analyze various production features for their ability to predict sow productivity in the subsequent parity, based on measurable traits recorded during the previous parity using machine learning models. Although these models demonstrated less than optimal performance on the datasets employed in this study, they showed considerable potential for enhancement with the inclusion of additional sow attributes and production records. In addition, these classifiers displayed strong generalizability on sow records from new farms, demonstrating significant potential for practical applications.

Read the full article in the Journal of Animal Science.