 |
International Society of Science and Applied Technologies |
| One-Class Classification: Advances, Applications, and Comparisons | ||||
| Author | Azadeh A. Zadeh
|
|||
| Co-Author(s) | Joffrey L. Leevy; Taghi M. Khoshgoftaar; Naeem Seliya
|
|||
| Abstract | In today’s data-driven era, there is an increasing demand to classify vast amounts of data, where the interest often lies in identifying the outlying, novel, or rare categories. This paper focuses on the dichotomy between binary classification and the more specialized one-class classification (OCC). Binary classification is widely used in numerous fields. However, challenges arise when there is a pronounced class imbalance or when one class is notably underrepresented. As a response to these challenges, OCC has emerged as a potent solution. We present a comprehensive review of four prominent OCC algorithms: One-Class Support Vector Machine (SVM), Support Vector Data Description (SVDD), Gaussian Mixture Model (GMM), and One- Class Generative Adversarial Network (GAN). Drawing from numerous studies, this paper underscores the effectiveness of these algorithms in diverse domains, benchmarks their performances based on dataset sizes and performance metrics, and presents scenarios where each algorithm is most aptly suited. The critical insight offered in this paper aims to guide researchers and practitioners in selecting the optimal classification method tailored to their specific requirements.
|
|||
| Keywords | machine learning, one-class classification, binary classification, class imbalance | |||
| Article #: RQD2024-149 | ||||
Proceedings of 29th ISSAT International Conference on Reliability & Quality in Design |