Classification of Milking Animals Based on Production and Reproductive Performance for Effective Management Decisions: An Approach Using K-Means Clustering

This study was conducted to determine the feasibility of implementing principal component analysis (PCA) and K-means clustering method for grouping cows in a dairy herd based on their performance on multiple traits related to milk production and reproduction, for implementing need based, group specific management practices. The proposed method was tested on two different management systems, namely System A (Friesian x Jersey crossbreds managed intensively under Low Country Dry Zone) and System B (Sahiwal crossbreds managed semi intensively under Low Country Intermediate Zone). Six large scale state dairy farms were grouped into the two systems with 1507 lactation records of 643 cows belonging to System A and 552 lactation records of 185 cows in System B. Means of calf birth weight, age at first calving, lactation length (LL), total milk production (MP), average daily yield, and calving interval (CI) of System A were significantly superior to the respective means of System B (p<0.05). The PCA used MP, LL and CI traits where all three traits exhibited significant positive pairwise correlations (p<0.05). For both systems, the first two PCs explained over 85% of the total variation of traits. Traits MP and LL played a major component in PC1 (production component) and CI was loaded heavily for PC2 (reproduction component). For both systems, the K- means clustering approach was effective with the three clusters of cows formed within a system showing significant differences for MP, LL and CI (p<0.05).