Prediction of Egg Weight in Japanese Quails (Coturnix coturnix japonica) with Internal Quality Traits Using Machine Learning Algorithms

Abstract

This study comparatively evaluates the performance of various machine learning (ML) algorithms to predict quail egg weight based on internal quality characteristics. Two hundred Japanese quail eggs were used, and measurements were taken for albumen length, height, and width, as well as yolk diameter, albumen index, yolk index, and Haugh unit. Linear Regression, Support Vector Machines (SVM), Random Forest, Gradient Boosting, and XGBoost algorithms were applied for egg weight prediction, with hyperparameter optimization performed using the GridSearchCV method. Performance was assessed using R2, RMSE, MAE, MAPE, RAE, and MAD metrics. SVM and Linear Regression models demonstrated high generalization ability by providing balanced results between training and test datasets. The SVM and Linear Regression achieved the highest R2 (0.990) and the lowest error values on the test data, making them the most successful algorithms. In contrast, XGBoost, Gradient Boosting, and Random Forest models showed high accuracy on training data, but experienced noticeable performance drops on test data, indicating overfitting issues. As a result, the SVM and linear regression models stand out as practical, reliable, and applicable methods in innovative farming applications aimed at increasing production efficiency, optimizing resource use, and establishing early decision support systems