Comparison of K-Means and Hierarchical Clustering Methods for Buffalo Milk Production Data

Area of Science:

• Animal Science
• Data Science
• Agricultural Management

Background:

• Effective herd management relies on understanding animal variations.
• Clustering algorithms offer potential for data-driven grouping of livestock.

Purpose of the Study:

• To evaluate K-means and hierarchical clustering for grouping Italian Mediterranean buffalo.
• To determine if data-driven groupings can improve herd management strategies.

Main Methods:

• Utilized routinely collected test-day records from three Italian Mediterranean buffalo herds.
• Applied K-means and hierarchical clustering algorithms to combined and individual herd datasets.
• Assessed clustering performance using silhouette scores, Davies-Bouldin Index (DBI), and Calinski-Harabasz Index (CHI).

Main Results:

• K-means clustering consistently outperformed hierarchical clustering across all datasets.
• K-means identified two clusters in most datasets, with one herd yielding three clusters.
• Days in milk, milk yield, age, and lactation stage were key factors differentiating clusters.

Conclusions:

• K-means clustering is a superior method for grouping buffalo based on test-day data.
• Identified animal subgroups can inform targeted herd management practices.
• This approach supports data-driven decision-making for improved buffalo production.