الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 155 |  |  |
| | | from | 155 | | |
Abstract
Milk yield optimization is critical for the economic sustainability of
dairy farms. This study investigates the potential of some classification
models, including linear discriminant analysis (LDA), flexible
discriminant analysis (FDA), ordinal logistic regression (OLR), and
artificial neural networks (ANNs) in predicting milk yield levels. Using
3793 lactation records from cows calved between 2009 and 2020,
researchers investigated some predictors such as age at first calving
(AFC), lactation order (LO), days open (DO), days in milk (DIM), dry
period (DP), calving season (CFS), 305- milk yield (305-MY), calving
interval (CI), and total services per conception (S/C). Model
performance was assessed using criteria such as overall accuracy,
sensitivity, specificity, positive predictive value, negative predictive
value, F1 score, and area under curve value (AUC). The predictors’
significance demonstrated that all the investigated parameters
statistically (P < 0.05) contributed to milk yield prediction, with 305-
MY, parities, and calving season being the most important
characteristics. Furthermore, the current findings revealed that
classification models were effective at predicting milk yield. ANNs
achieved the best accuracy (0.94), with an AUC of 0.92, then followed
by FDA, LDA, and OLR. In conclusion, classification models are a
useful tool for accurate and efficient milk yield prediction in dairy farms.
This will enable farmers to understand milk yield determinants, optimize
production, enhance farm performance, and contribute to animal welfare
through informed decisions.