Şenol Çelik, Turgay Şengül, Ömer Şengül
Significant effects of genotype and year were observed for days to pollen shedding, plant height, and ear height. Grain moisture was significantly influenced by year and the interaction between genotype and year, whereas grain yield was primarily affected by genotype. Among the tested genotypes, P2088 yielded the highest average grain yield (14.50 t ha-1), while KALUMET showed the latest flowering and the tallest plants. Protein content was the only compositional trait significantly impacted by genotype, with DKC6680 showing the highest average protein content (10.33%) and P2088 the lowest (9.16%). Conversely, mineral traits remained statistically stable across genotypes and years.
INTRODUCTION: This study aims to evaluate the predictive performance of different machine learning algorithms and to identify the key factors influencing eggshell breaking resistance in laying hens.
METHODS: Three machine learning methods, C5.0 decision tree, Random Forest (RF), and Support Vector Regression (SVR), were applied to egg classification and prediction tasks. Eggshell strength (ER) was predicted using egg weight (EW), shell weight (SW), shell thickness (ST), and shape index (SI).
RESULTS: The C5.0 decision tree, trained on 464 eggs, achieved an overall classification accuracy of 65.1%, with a tendency to misclassify brown eggs as white, suggesting potential feature overlap or class imbalance. Among the regression models, RF outperformed SVR, yielding higher R2 (0.852 vs. 0.553) and adjusted R2 (0.835 vs. 0.548) values, along with lower error metrics (MSE, RMSE, and MAPE). In addition to superior predictive accuracy, the RF model provided insights into the relative importance of egg quality traits affecting eggshell breaking resistance.
DISCUSSION: Overall, the findings indicate that ensemble-based machine learning methods are effective tools for both accurate prediction and identification of influential factors related to eggshell strength.