LOF-enhanced SMOTE algorithm for imbalanced dataset

Zhuangzhuang Zhang; Jing Hu; Tiecheng Song

doi:10.1117/12.2685807

15 August 2023 LOF-enhanced SMOTE algorithm for imbalanced dataset

Zhuangzhuang Zhang, Jing Hu, Tiecheng Song

Proceedings Volume 12719, Second International Conference on Electronic Information Technology (EIT 2023); 127193G (2023) https://doi.org/10.1117/12.2685807
Event: Second International Conference on Electronic Information Technology (EIT 2023), 2023, Wuhan, China

Abstract

This paper proposes a new algorithm, LOF-Enhanced SMOTE, aimed at addressing the problem of imbalanced datasets in machine learning tasks. Due to the significantly fewer samples of certain classes in imbalanced datasets, the performance of classifiers may be negatively affected. To solve this problem, we introduce the Local Outlier Factor (LOF) algorithm to remove boundary noise on the basis of the SMOTE algorithm, and use a Gaussian kernel function to consider the similarity of generated samples. We conduct experiments on real intrusion detection data, UNSW-NB15. The results show that LOF-Enhanced SMOTE outperforms SMOTE and Borderline-SMOTE algorithms overall, and significantly outperforms them in detecting certain minority classes. This indicates that the LOF-Enhanced SMOTE algorithm can effectively solve the classification problem of imbalanced datasets.

Citation Download Citation

Zhuangzhuang Zhang, Jing Hu, and Tiecheng Song "LOF-enhanced SMOTE algorithm for imbalanced dataset", Proc. SPIE 12719, Second International Conference on Electronic Information Technology (EIT 2023), 127193G (15 August 2023); https://doi.org/10.1117/12.2685807

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available