Standardized Sizing Through Machine Learning (SML): Utilizing Decision Tree Algorithm to Classify White Chicken Egg Sizes

This study presents a machine learning approach for standardization of sizing, aiding the subjectivity introduced by traditional method used by small-scale farms due to lack of accessibility to such technologies. Features including egg weight, and physical dimensions of white chicken eggs retrieved from 628 samples, equally divided into four size groups. Multiple machine learning models were trained and evaluated. Notably, the Decision Tree algorithm demonstrated the highest accuracy during training at 99.8%, outperforming other algorithms such as Gaussian Naïve Bayes (94.8%), Bernoulli Naïve Bayes (95.4%), Linear Discriminant Analysis (96.4%), K-nearest neighbors (98.2%), Logistic Regression (99.2%), Random Forest (99.8%), Adaptive Boosting (99.8%), and Gradient Boost (99.8%). During testing, the five adamant models were chosen to test with Decision Tree still becoming as prominent with AdaBoost and Gradient Boost, showing a 98.4% accuracy, with Random Forest tailing at 97.7% and lastly is Support Vector Classifier with 92.97%. With Decision Tree, its robust performance establishes the algorithm as the optimal algorithm for predicting egg size. Compared to inaccurate human procedures, the model offers an objective, uniform, and optimal system for size grading. This can improve local farms' marketability and profitability while also making it easier to comply with national size standards.