An application for the classification of egg quality and haugh unit based on characteristic egg features using machine learning models

Abstract

With the increase in the world population, the nutritional needs of people have been increased. The demand for eggs which is one of the most important food sources has been increased over years. Therefore, it is very important to inform people about egg quality in order to be prepared for adverse situations such as substitution, mislabeling, and fraud. In this study, it is aimed to specify egg quality without using haugh unit (HU). Besides, another aim is to find how much information HU carries about the specification of egg quality. A dataset including 20 features related to eggs taken from 438 chickens created by Poultry Research Institute (PRI) has been analyzed. An application that can classify egg qualities as very good and excellent using machine learning (ML) models like decision tree (DT), linear discriminant analysis (LDA), logistic regression (LR), naïve bayes (NB), support vector machines (SVM), K-nearest neighboring (KNN), random forest (RF) and artificial neural networks (ANN) has been developed in this study. In addition to that, HU at the 24th week and the 32nd week which are the most important classifier to determine egg qualities have been classified as low informative, medium informative, and high informative by the developed application. Egg quality has the best been classified by LR model based on accuracy and Matthews correlation coefficient (MCC) values as 98.6% and 0.96, respectively. HU at the 24th week has the best been classified by RF based on accuracy and MCC as 96.8% and 0.93, respectively. HU at the 32nd week has the best been classified by RF based on accuracy and MCC as 95.1% and 0.92, respectively. This paper mainly focuses on the classification of egg quality based on not only HU but also egg characteristic features and the importance of the informative feature of HU to classify egg quality.