Machine Learning vs Traditional Methods: A New Look at Predicting Broiler Traits

Machine learning (ML) is quickly becoming a big deal in lots of areas, from research to practical things like predicting how well livestock will breed. But not many people have looked into using ML to improve chicken breeding, especially for yellow-feathered broilers. This study tried out seven different ML methods – support vector regression (SVR), random forest (RF), gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), kernel ridge regression (KRR), and multilayer perceptron (MLP) – to see if they could better predict how well these chickens would lay eggs and grow.

The old ways, like GBLUP and Bayesian methods, still turned out to be more accurate for most traits. But when it came to predicting half-eviscerated weight (HEW), ML methods showed a big improvement, up to 61. 3% better than the old ways. For eviscerated weight (EW), MLP was the winner, with a 19% improvement. Interestingly, tweaking the settings (hyperparameters) of these ML methods helped even more. The average improvement in prediction accuracy ranged from 13. 2% to 46. 3% when they were adjusted. And here's a cool thing: using a genome-wide association study (GWAS) to pick out important genetic markers could make these predictions even better. This study is a great way to see how ML can be used in chicken breeding. It's like having a new tool in the toolbox to help make better, healthier chickens.

Actions