Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Confronting Discrimination in Classification: Smote Based on Marginalized Minorities in the Kernel Space for Imbalanced Data

Lingyun Zhong

TL;DR

The paper addresses discrimination against minority fraud samples in highly imbalanced classification settings. It introduces MM-SMOTE, an adaptive oversampling framework guided by a basic SVM that identifies minority support vectors, weights them by their distance to the decision boundary using $L(x)$, and generates synthetic samples in the kernel space, followed by kernel-matrix augmentation to retrain a new classifier. Key contributions include a principled minority-sample selection mechanism, density-aware adaptive oversampling, and a unified kernel-space augmentation that improves minority detection. Experiments on the Kaggle Credit Card Fraud Detection dataset show that MM-SMOTE delivers superior F1-score and G-mean and demonstrates stability across a range of imbalance ratios, highlighting its potential for robust discrimination in financial fraud tasks.

Abstract

Financial fraud detection poses a typical challenge characterized by class imbalance, where instances of fraud are extremely rare but can lead to unpredictable economic losses if misidentified. Precisely classifying these critical minority samples represents a challenging task within the classification. The primary difficulty arises from mainstream classifiers, which often exhibit "implicit discrimination" against minority samples in evaluation metrics, which results in frequent misclassifications, and the key to the problem lies in the overlap of feature spaces between majority and minority samples. To address these challenges, oversampling is a feasible solution, yet current classical oversampling methods often lack the necessary caution in sample selection, exacerbating feature space overlap. In response, we propose a novel classification oversampling approach based on the decision boundary and sample proximity relationships. This method carefully considers the distance between critical samples and the decision hyperplane, as well as the density of surrounding samples, resulting in an adaptive oversampling strategy in the kernel space. Finally, we test the proposed method on a classic financial fraud dataset, and the results show that our proposed method provides an effective and robust solution that can improve the classification accuracy of minorities.

Confronting Discrimination in Classification: Smote Based on Marginalized Minorities in the Kernel Space for Imbalanced Data

TL;DR

The paper addresses discrimination against minority fraud samples in highly imbalanced classification settings. It introduces MM-SMOTE, an adaptive oversampling framework guided by a basic SVM that identifies minority support vectors, weights them by their distance to the decision boundary using , and generates synthetic samples in the kernel space, followed by kernel-matrix augmentation to retrain a new classifier. Key contributions include a principled minority-sample selection mechanism, density-aware adaptive oversampling, and a unified kernel-space augmentation that improves minority detection. Experiments on the Kaggle Credit Card Fraud Detection dataset show that MM-SMOTE delivers superior F1-score and G-mean and demonstrates stability across a range of imbalance ratios, highlighting its potential for robust discrimination in financial fraud tasks.

Abstract

Financial fraud detection poses a typical challenge characterized by class imbalance, where instances of fraud are extremely rare but can lead to unpredictable economic losses if misidentified. Precisely classifying these critical minority samples represents a challenging task within the classification. The primary difficulty arises from mainstream classifiers, which often exhibit "implicit discrimination" against minority samples in evaluation metrics, which results in frequent misclassifications, and the key to the problem lies in the overlap of feature spaces between majority and minority samples. To address these challenges, oversampling is a feasible solution, yet current classical oversampling methods often lack the necessary caution in sample selection, exacerbating feature space overlap. In response, we propose a novel classification oversampling approach based on the decision boundary and sample proximity relationships. This method carefully considers the distance between critical samples and the decision hyperplane, as well as the density of surrounding samples, resulting in an adaptive oversampling strategy in the kernel space. Finally, we test the proposed method on a classic financial fraud dataset, and the results show that our proposed method provides an effective and robust solution that can improve the classification accuracy of minorities.
Paper Structure (15 sections, 10 equations, 2 figures, 6 tables)

This paper contains 15 sections, 10 equations, 2 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Imbalanced Classification at Data Preprocessing Level
  4. Imbalanced Classification at Algorithm Level
  5. Methodology
  6. Key Sample Identification
  7. Key Sample Selection
  8. Synthetic Sample Generation
  9. Kernel Matrix Augment
  10. Experiments
  11. Dataset
  12. Baseline Methods
  13. Experimental Settings
  14. Result Anaysis
  15. Conclusion

Figures (2)

  • Figure 1: SVM classification samples. $H_0$ represents the decision hyperplane;$H_1$ and $H_2$ represent hyperplanes generalized by majorities and minorities. The points represented by blue triangles are safe points that are correctly classified among minorities. Points located in the yellow margin area and its boundaries are points that may be misclassified; red points represent points that will definitely be misclassified. , the latter two points are support vectors, which are also the focus of this study.
  • Figure 2: 3 cases of support vectors and their $k$ nearest neighbors. The red triangles represent the selected sampling points, the blue triangles represent the surrounding minority class samples, and the green squares represent the surrounding majority class samples.