Related Experiment Video
Updated: Sep 16, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.9K
Machine Learning Approaches for the Prediction of Displaced Abomasum in Dairy Cows Using a Highly Imbalanced Dataset
Zeinab Asgari1, Ali Sadeghi-Sefidmazgi2, Abbas Pakdel1
1Department of Animal Sciences, College of Agriculture, Isfahan University of Technology, Isfahan 84156-83111, Iran.
Animals : an Open Access Journal From MDPI
|July 12, 2025
Summary
Early prediction of displaced abomasum (DA) in dairy cows is crucial for reducing economic losses. Machine learning models, particularly Gradient Boosting Machines (GBM) and Random Forest (RF), show promise in identifying susceptible cows.
Area of Science:
- Veterinary Medicine
- Animal Science
- Machine Learning Applications
Background:
- Displaced abomasum (DA) causes significant economic losses in dairy farming due to decreased milk yield and premature culling.
- Early identification of cows susceptible to DA is vital for effective management and mitigation of financial impacts.
- DA is a complex trait with a low incidence, making prediction challenging.
Purpose of the Study:
- To investigate the predictive potential of machine learning algorithms for early detection of displaced abomasum (DA) in Holstein dairy cows.
- To evaluate the performance of Logistic Regression, Naïve Bayes, Decision Tree, Random Forest, and Gradient Boosting Machines in predicting DA.
- To identify reliable models that can assist dairy farmers in proactive management strategies.
Main Methods:
- Utilized 20 herd-cow-specific features and sire genetic information from 7 Holstein dairy herds (calving 2010-2020).
- Compared the efficacy of five machine learning algorithms: Logistic Regression (LR), Naïve Bayes (NB), Decision Tree, Random Forest (RF), and Gradient Boosting Machines (GBM).
- Assessed model performance using metrics such as F2 score and true positive rate, particularly on imbalanced data.
Main Results:
- Gradient Boosting Machines (GBM) and Random Forest (RF) demonstrated superior performance in predicting DA, achieving an F2 measure of 0.32.
- The Random Forest (RF) model exhibited the highest true positive rate at 0.75, followed closely by GBM at 0.70.
- Despite highly imbalanced data, the study successfully showed the potential for forecasting DA-susceptible cases.
Conclusions:
- Machine learning models, especially RF and GBM, offer a viable approach for the early prediction of displaced abomasum in dairy cattle.
- This predictive capability can empower dairy farmers with data-driven insights for timely management interventions.
- Implementing such prediction tools can help minimize economic losses associated with DA in dairy herds.
