Machine learning approaches for the prediction of retained placenta in dairy cows.

Journal: Theriogenology
Published Date:

Abstract

Retained placenta (RP) is a reproductive disorder that causes significant financial losses to the dairy industry. Predicting RP risk in cows post-calving is a challenging task. This study aimed to evaluate the predictive capabilities of five machine learning algorithms Naïve Bayes (NB), Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), and XGBoost along with Logistic Regression (LR) in predicting RP incidence using data from 363,945 calving records (72,788 affected and 291,092 unaffected) and 16 predictor features from 15 dairy herds in Iran. The performance of these algorithms was assessed based on key metrics, including the area under the receiver operating characteristic curve (AUC), F1-score, and accuracy. The results showed that XGBoost (AUC 0.78) and RF (AUC 0.78) significantly outperformed other algorithms, while XGBoost achieved the highest F1-score (41 %), indicating its potential for reliable RP prediction. Logistic Regression and Naïve Bayes had similar AUC values (0.66 and 0.67, respectively), suggesting they may be less effective for this task. Despite limitations such as missing environmental and management data, the study demonstrates the strong potential of machine learning models, particularly XGBoost, as a decision support tool for RP prediction and management in precision dairy farming.

Authors

  • Mohammad Rahimi Hosseinabadi
    Department of Animal Science, College of Agriculture, Isfahan University of Technology, Isfahan, 84156-83111, Iran.
  • Amir Hossein Mahdavi
    Department of Animal Science, College of Agriculture, Isfahan University of Technology, Isfahan, 84156-83111, Iran. Electronic address: mahdavi@iut.ac.ir.
  • Abolfazl Mahnani
    Arona Chemie P.J.S. Co., 15546-43656, Tehran, Iran.
  • Zeinab Asgari
    Department of Animal Science, College of Agriculture, Isfahan University of Technology, Isfahan, 84156-83111, Iran.
  • Saleh Shahinfar
    Department of Computer Science, School of Science and Technology, University of New England, Armidale, NSW, Australia. Electronic address: shahinfar@uwalumni.com.