A key-feature-based clustering approach to assess the impact of technology integration on cow health in Austrian dairy farms

Abstract

In the domain of precision livestock farming, the integration of diverse data sources is crucial for advancing sustainability and evaluating the implications of farm management practices on cow health. Addressing the challenge of data heterogeneity and management diversity, we propose a key-feature-based clustering method. This approach, merging knowledge-driven feature selection with unsupervised machine learning, enables the systematic investigation of management effects on cow health by forming distinct clusters for analysis. Utilizing data from 3,284 Austrian farms, including 80 features related to feeding, milking, housing, and technology systems, and health information for 56,000 cows, we show how this methodology can be applied to study the impact of technological systems on cow health resulting from the incidence of veterinary diagnoses. Our analysis successfully identified 14 distinct clusters, further divided into four main groups based on their level of technological integration in farm management: “SMART,” “TRADITIONAL,” “AMS (automatic milking system),” and “SENSOR.” We found that “SMART” farms, which integrate both AMS and sensor systems, exhibited a minimally higher disease risk for milk fever (OR 1.09) but lower risks for fertility disorders and udder diseases, indicating a general trend toward reduced disease risks. In contrast, farms with “TRADITIONAL” management, without AMS and sensor systems, showed the lowest risk for milk fever but the highest risk of udder disease (OR 1.12) and a minimally higher incidence of fertility disorders (OR 1.07). Furthermore, across all four groups, we observed that organic farming practices were associated with a reduced incidence of milk fever, udder issues, and particularly fertility diagnoses. However, the size of the effect varied by cluster, highlighting the complex and multifactorial nature of the relationship between farm management practices and disease risk. The study highlights the effectiveness of the key-feature-based clustering approach for high-dimensional data analyses aimed at comparing different management practices and exploring their complex relationships. The adaptable analytical framework of this approach makes it a promising tool for planning optimizing sustainable and efficient animal husbandry practices.