Oversampling technique based on fuzzy representativeness difference for classifying imbalanced data

Ruonan Ren

doi:10.1007/s10489-020-01644-0

Return

Oversampling technique based on fuzzy representativeness difference for classifying imbalanced data

2020-03-10

PRE

Ruonan Ren *

Youlong Yang

Liqin Sun

DOI：10.1007/s10489-020-01644-0

Original

Original request for help

Save

Abstract

En 中文

Class imbalance problem poses a difficulty to learning algorithms in pattern classification. Oversampling techniques is one of the most widely used techniques to solve these problems, but the majority of them use the sample size ratio as an imbalanced standard. This paper proposes a fuzzy representativeness difference-based oversampling technique, using affinity propagation and the chromosome theory of inheritance (FRDOAC). The fuzzy representativeness difference (FRD) is adopted as a new imbalance metric, which focuses on the importance of samples rather than the number. FRDOAC firstly finds the representative samples of each class according to affinity propagation. Secondly, fuzzy representativeness of every sample is calculated by the Mahalanobis distance. Finally, synthetic positive samples are generated by the chromosome theory of inheritance until the fuzzy representativeness difference of two classes is small. A thorough experimental study on 16 benchmark datasets was performed and the results show that our method is better than other advanced imbalanced classification algorithms in terms of various evaluation metrics.

Keywords:

Class imbalanced problem

Oversampling technique

Affinity propagation

Chromosome theory of inheritance

Fuzzy representativeness difference

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.