Comparative Evaluation of Machine Learning Algorithms for Predicting Body Weight from Linear Body Measurements in Zambian Indigenous Chickens

Machine learning has not been widely applied to Zambian indigenous chickens, which account for most of the poultry kept by smallholder farmers in the country. We compared six algorithms (CART, ANN, KNN, Random Forest, SVR, and Gradient Boosting) to determine which best predicted body weight from eight linear body measurements. 618 chickens were phenotyped in accordance with FAO (2012) guidelines, with body weight recorded on a digital scale and linear measurements taken with a tailor’s tape. The data were split 80:20 into training (n = 496) and test (n = 122) sets, tuned via repeated 10-fold cross-validation (5 repeats), and performance on the test set was assessed using six goodness-of-fit criteria. Random Forest was the most accurate according to the numerical metrics (R2 = 0.816, RMSE = 0.204 kg), followed by Gradient Boosting (R2 = 0.789) and SVR (R2 = 0.766). A paired Wilcoxon test showed that Random Forest was significantly better than all other algorithms, including Gradient Boosting (adjusted p < 0.001). Generalization was assessed by comparing the cross-validated RMSE to the test RMSE. All six algorithms had test-to-CV ratios between 0.90 and 1.22. The pruned CART tree generalised best, and ANN had the largest gap despite having the lowest test accuracy. Chest circumference had the highest correlation with body weight (r = 0.696) and ranked first in permutation importance for all six algorithms (no exceptions). A single chest circumference measurement in a linear regression accounted for 45.4% of the variance and thus provides a useful field estimate where scales are not available. The full models, particularly Random Forest, were substantially more accurate and are recommended for predicting body weight in Zambian indigenous chickens. Chest circumference is the key practical trait for selection and management programs.

Authors

Institutions

Publication Details

Journal
Animals
Published
2026-11-12
DOI
https://doi.org/10.3390/ani16223576
Primary Topic
Livestock and Poultry Management
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Comparative Evaluation of Machine Learning Algorithms for Predicting Body Weight from Linear Body Measurements in Zambian Indigenous Chickens

Simushi Liswaniso, Thobela Louis Tyasi, Salome Mhlabini, Prudence Shibe Mokgapa et al.
Animals
Livestock and Poultry Management
article

Comparative Evaluation of Machine Learning Algorithms for Predicting Body Weight from Linear Body Measurements in Zambian Indigenous Chickens

Simushi Liswaniso, Thobela Louis Tyasi, Salome Mhlabini, Prudence Shibe Mokgapa, Jinghua Zhao
article en

Abstract

Machine learning has not been widely applied to Zambian indigenous chickens, which account for most of the poultry kept by smallholder farmers in the country. We compared six algorithms (CART, ANN, KNN, Random Forest, SVR, and Gradient Boosting) to determine which best predicted body weight from eight linear body measurements. 618 chickens were phenotyped in accordance with FAO (2012) guidelines, with body weight recorded on a digital scale and linear measurements taken with a tailor’s tape. The data were split 80:20 into training (n = 496) and test (n = 122) sets, tuned via repeated 10-fold cross-validation (5 repeats), and performance on the test set was assessed using six goodness-of-fit criteria. Random Forest was the most accurate according to the numerical metrics (R2 = 0.816, RMSE = 0.204 kg), followed by Gradient Boosting (R2 = 0.789) and SVR (R2 = 0.766). A paired Wilcoxon test showed that Random Forest was significantly better than all other algorithms, including Gradient Boosting (adjusted p < 0.001). Generalization was assessed by comparing the cross-validated RMSE to the test RMSE. All six algorithms had test-to-CV ratios between 0.90 and 1.22. The pruned CART tree generalised best, and ANN had the largest gap despite having the lowest test accuracy. Chest circumference had the highest correlation with body weight (r = 0.696) and ranked first in permutation importance for all six algorithms (no exceptions). A single chest circumference measurement in a linear regression accounted for 45.4% of the variance and thus provides a useful field estimate where scales are not available. The full models, particularly Random Forest, were substantially more accurate and are recommended for predicting body weight in Zambian indigenous chickens. Chest circumference is the key practical trait for selection and management programs.

AnimalsVol. 16(22)
Changchun Normal University (CN), University of Zambia (ZM), University of Limpopo (ZA)
Zero hunger
Livestock and Poultry Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon