Quality of performance evaluation of ten machine learning algorithms in classifying thirteen types of apple fruits

Abstract

Recently, computer vision technology has become essential for the automatic, accurate, and fast classification of fruits. Actually, there are many challenges in separating the types of fruits that are somewhat similar, such as apples, pears, and peaches. However, the challenges become more difficult if the separation is on different varieties of the same fruit. While the difficulty doubles if the classification takes place with a large number of different varieties of the same fruit. Most of the literature which is presented in this regard, and which is relied on the use of machine learning techniques lacked the following: first; the focus was on certain technologies such as k-nearest neighbor (KNN), support vector machine (SVM) without looking at many other machine learning techniques. Second; the literature was concerned only with measuring the accuracy of the techniques that are used, without looking at the relationship between the accuracy and processing speed (computation times). This manuscript aims to study and analyze the results of measuring accuracy and computation times for ten machine-learning techniques in order to identify and classify thirteen types of apples. After studying and analyzing the results, many observations were made, which will be referred to in the results section.