A theoretical distribution analysis of synthetic minority oversampling technique (SMOTE) for imbalanced learning

被引:65
作者
Elreedy, Dina [1 ]
Atiya, Amir F. [1 ]
Kamalov, Firuz [2 ]
机构
[1] Cairo Univ, Comp Engn Dept, Giza 12613, Egypt
[2] Canadian Univ Dubai, Dept Elect Engn, Dubai 117781, U Arab Emirates
关键词
SMOTE; Class imbalance; Distribution density; Over-sampling; Minority class; SAMPLING APPROACH; CLASSIFICATION;
D O I
10.1007/s10994-022-06296-4
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Class imbalance occurs when the class distribution is not equal. Namely, one class is under-represented (minority class), and the other class has significantly more samples in the data (majority class). The class imbalance problem is prevalent in many real world applications. Generally, the under-represented minority class is the class of interest. The synthetic minority over-sampling technique (SMOTE) method is considered the most prominent method for handling unbalanced data. The SMOTE method generates new synthetic data patterns by performing linear interpolation between minority class samples and their K nearest neighbors. However, the SMOTE generated patterns do not necessarily conform to the original minority class distribution. This paper develops a novel theoretical analysis of the SMOTE method by deriving the probability distribution of the SMOTE generated samples. To the best of our knowledge, this is the first work deriving a mathematical formulation for the SMOTE patterns' probability distribution. This allows us to compare the density of the generated samples with the true underlying class-conditional density, in order to assess how representative the generated samples are. The derived formula is verified by computing it on a number of densities versus densities computed and estimated empirically.
引用
收藏
页码:4903 / 4923
页数:21
相关论文
共 65 条
[1]  
Abd Elrahman S.M., 2013, J NETW INNOV COMPUT, V1, P332
[2]  
Ahsan M, 2018, INT CONF ELECTRO INF, P531, DOI 10.1109/EIT.2018.8500086
[3]  
Al-Sirehy F., 2013, INT J APPL MATH, V26, P191
[4]  
Al-Sirehy F., 2013, APPL MATH SCI, V7, P3489
[5]   The quest for the optimal class distribution: An approach for enhancing the effectiveness of learning via resampling methods for imbalanced data sets [J].
Albisua I. ;
Arbelaitz O. ;
Gurrutxaga I. ;
Lasarguren A. ;
Muguerza J. ;
Pérez J.M. .
Pérez, J. M. (txus.perez@ehu.es), 1600, Springer Verlag (02) :45-63
[6]  
Atiya A, 1997, 1997 IEEE INTERNATIONAL CONFERENCE ON NEURAL NETWORKS, VOLS 1-4, P2112, DOI 10.1109/ICNN.1997.614231
[7]   SMOTE-Based Homogeneous Ensemble Methods for Software Defect Prediction [J].
Balogun, Abdullateef O. ;
Lafenwa-Balogun, Fatimah B. ;
Mojeed, Hammed A. ;
Adeyemo, Victor E. ;
Akande, Oluwatobi N. ;
Akintola, Abimbola G. ;
Bajeh, Amos O. ;
Usman-Hamza, Fatimah E. .
COMPUTATIONAL SCIENCE AND ITS APPLICATIONS - ICCSA 2020, PT VI, 2020, 12254 :615-631
[8]   Strategies for learning in class imbalance problems [J].
Barandela, R ;
Sánchez, JS ;
García, V ;
Rangel, E .
PATTERN RECOGNITION, 2003, 36 (03) :849-851
[9]  
Batista G.E., 2004, ACM SIGKDD EXPL NEWS, V6, P20, DOI [10.1145/1007730.1007735, 10.1145/1007730.1007735.2, DOI 10.1145/1007730.1007735]
[10]   LoRAS: an oversampling approach for imbalanced datasets [J].
Bej, Saptarshi ;
Davtyan, Narek ;
Wolfien, Markus ;
Nassar, Mariam ;
Wolkenhauer, Olaf .
MACHINE LEARNING, 2021, 110 (02) :279-301