AI-driven prediction of body weight in chicken genotypes with different growth rates
This study evaluates four machine learning algorithms: Gaussian Exponential Regression (GER), Feedforward Neural Networks (FFNN), Support Vector Machines (SVM), and Decision Trees (DT) for predicting body weight in chickens with different growth rates (slow-, medium-, and fast-growing breeds). It fi...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Springer Science and Business Media B.V.
2026
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/123310/1/123310.pdf http://psasir.upm.edu.my/id/eprint/123310/ https://link.springer.com/article/10.1007/s11250-026-04925-x |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study evaluates four machine learning algorithms: Gaussian Exponential Regression (GER), Feedforward Neural Networks (FFNN), Support Vector Machines (SVM), and Decision Trees (DT) for predicting body weight in chickens with different growth rates (slow-, medium-, and fast-growing breeds). It fills a major research gap by benchmarking these models under uniform conditions across multiple growth categories. Starting at 14 days of age, 300 male chicks (100 of the genotypes SAGA, Sasso, and Cobb 500, which represent slow-, medium-, and fast-growing breeds, respectively) were used in a controlled experiment. Weekly body weight and six (6) morphometric traits were recorded for four weeks and split into training (75%) and validation (25%) sets. A total of 2,160 data records were obtained from weekly morphometric measurements across 300 birds. Model performance was measured by correlation coefficient (R), coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). Across all growth groups, GER consistently performed better than other models, exhibiting the highest accuracy and lowest errors. FFNN performed well for slow- and medium-growing breeds but showed increased errors for fast-growing Cobb 500. SVM and DT had the poorest results, with higher errors and poor generalizability. Model accuracy was ranked as follows: GER > FFNN > DT/SVM. This study provides recommendations for selecting machine learning models tailored to specific chicken growth rates for accurate weight prediction. Future work should validate these findings on commercial farms and explore hybrid models to improve robustness. GER’s superior performance highlights its potential as a reliable and efficient tool for precision poultry farming. |
|---|