The Application of Ordinal Logistic Regression Model as a Robust Tool for Enhanced Prediction of Milk Yield in Dairy Cows

Abstract

Milk yield is a vital issue of concern for dairy cows. Hence, accurate milk production prediction is critical for improving dairy farm management and profitability. The purpose of this study was to examine the feasibility of applying ordinal logistic regression (OLR) to classify and predict milk production in Friesian cows into low (4500 kg), moderate (4500-7500 kg), and high (>7500 kg) classes. The data includes 3793 lactation records from dairy cows calved between 2009 and 2020 to investigate several explanatory variables, including the 305-day milk yield (305-MY), age at first calving (AFC), calving interval (CI), calving season (CFS), days open (DO), days in milk (DIM), dry period (DP), lactation order (LO), and number of services per conception (SPC). Significant determinants impacting yield were found, with varying impacts across different yield classes. The results suggested that LO, DIM, and 305-MY were the most significant parameters (P < 0.05) influencing data categorization. The OLR model demonstrated a satisfactory fit in predicting milk yield categories, as it showed considerable accuracy (56%) and an area under the curve equal to 0.69. In conclusion, the ordinal logistic regression demonstrated to be an effective method for modeling milk production as an ordinal parameter. The model's results provide insights into the complex interaction of factors influencing milk output, and directing management strategies for optimal production.