Related Experiment Video
Updated: Apr 14, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.7K
Weaning performance prediction in lactating sows using machine learning, for precision nutrition and intelligent
Jiayi Su1, Xiangfeng Kong2, Wenliang Wang1
1Key Laboratory of Hunan Province for the Products Quality Regulation of Livestock and Poultry, College of Animal Science and Technology, Hunan Agricultural University, Changsha 410128, China.
Animal Nutrition (Zhongguo Xu Mu Shou Yi Xue Hui)
|June 9, 2025
Summary
Precision feeding models accurately predict swine weaning performance, improving farm sustainability and profitability. Key factors identified include lactation duration and birth litter weight for optimized swine production.
Area of Science:
- Animal Science
- Agricultural Engineering
- Data Science
Background:
- Traditional swine lactation feeding strategies risk nutrient deficiencies, impacting long-term productivity, sustainability, and profitability.
- Precision feeding, utilizing advanced predictive models, offers a data-driven approach to optimize swine production efficiency.
Purpose of the Study:
- Develop robust prediction models for key swine weaning performance indicators: weaned litter weight (WLW), weaned litter size (WLS), dry matter in milk (DMm), and nitrogen in milk (Nm).
- Integrate farm management practices and feed nutrient composition into a predictive framework for enhanced swine production outcomes.
Main Methods:
- Collected 10,089 observations from 17 pig farms across China.
- Employed 11 statistical and machine learning (ML) regression algorithms, including stratified sampling and recursive feature elimination for feature selection.
- Utilized Shapley Additive Explanations (SHAP) for feature importance analysis.
Main Results:
- Ensemble learning models (Random Forest, Gradient Boosting Decision Tree) demonstrated superior performance (R²: 0.40–0.80, MAE: 0.11–4.36).
- Consistently important predictors identified across models include lactation duration, birth litter weight, parity, and day 7 lactation backfat thickness (L.d7BF).
- Discrepancies in feature importance highlight non-linear relationships and feature interactions within the data.
Conclusions:
- The developed prediction models offer a novel framework for understanding and optimizing swine weaning performance.
- Identified key predictors provide actionable insights for improving swine production efficiency and sustainability.
- Optimized models can guide real-time, sensor-based precision feeding systems for enhanced swine farming outcomes.
