Document Type
Article
Publication Date
Fall 11-6-2024
Abstract
Data mining; known as data dredging or data archeology; provides several techniques to extract new and interpretable information from existing datasets. This work presents a comparison of the performance of various classification techniques, specifically Support Vector Machine (SVM), Radial Basis Function (RBF), K-Nearest Neighbors (KNN), and decision trees, using images of fruits. An open-source dataset called fruits360 is used focusing on three specific fruits: pineapple, cocos, and avocado. This selection allows testing the classification techniques in different ways, as pineapples and cocos are similar in color, while cocos and avocados resemble each other in shape due to their oval-like geometry. Among the classifiers tested, the KNN classifier achieved the highest accuracy score of 97.9%. Additionally, at the end of the paper, the impact of Principal Component Analysis (PCA), one of the most renowned dimensionality reduction techniques, on the performance of the classification tasks is examined.
Recommended Citation
F. Elkabbany, Ghada; G. Fouad, Youssef; W. Youssef, Kerollos; A. Amir, Jannah; and Zayed, Nourhan, "A Comparative Analysis of Optimizing Classification Techniques on Fruits Images Dataset" (2024). Mechanical Engineering. 132.
https://buescholar.bue.edu.eg/mech_eng/132
