Classification of broiler chicken eggs using support vector machine (svm) and feature selection algorithm

Abstract

According to the National Standardization Agency of Indonesia, consumed chicken eggs are classified based on their eggshell color and weights. This research aimed to incorporate computer vision and machine learning technology to eggs’ categorization process as an alternative to the standard and manual method. We used Hue Saturation Value (HSV) to store the eggs’ color space and Support Vector Machine (SVM) as the classification algorithm because of its robustness in learning data. A feature selection algorithm, Wrapper, was also applied to increase classification accuracy. The dataset used consists of 60 egg data with eight noted attributes (four of numeric type and four of nominal type with the last attribute as the class): H-value, S-value, V-value, weight, color, density, area, and weight class. The feature selection algorithm evaluated a total number of 29 subsets and found one subset as the candidate, consisting of only one attribute: Area. There were six support vectors found, and the coefficients of the vectors were: 1, 0.668, 0.334, 0.1289, 0.0684, and 0.4688. The classification results with three experiment scenarios have accuracy values of 100%, which was an improvement of the result of the previous work by the authors. This shows that SVM is a good and robust algorithm for classification.