DBSMOTE: Density-Based Synthetic Minority Over-sampling TEchnique

Abstract

A dataset exhibits the class imbalance problem when a target class has a very small number of instances relative to other classes. A trivial classifier typically fails to detect a minority class due to its extremely low incidence rate. In this paper, a new over-sampling technique called DBSMOTE is proposed. Our technique relies on a density-based notion of clusters and is designed to oversample an arbitrarily shaped cluster discovered by DB-SCAN. DBSMOTE generates synthetic instances along a shortest path from each positive instance to a pseudocentroid of a minority-class cluster. Consequently, these synthetic instances are dense near this centroid and are sparse far from this centroid. Our experimental results show that DBSMOTE improves precision, F-value, and AUC more effectively than SMOTE, Borderline-SMOTE, and Safe-Level-SMOTE for imbalanced datasets.

An insight into imbalanced Big Data classification: outcomes and challenges

“…On the other hand, it can result in the over-generalization problem. To tackle this issue of over-generalization versus increasing the minority class representation, various adaptive sampling methods have been proposed in recent years [29][30][31].…”

Improved classification of large imbalanced data sets using rationalized technique: Updated Class Purity Maximization Over_Sampling Technique (UCPMOT)

“…It helps to address the data features of small disjuncts and lack of density. Moreover it compiles the objectives [41] of elevating centroids based over_sampling.…”

Probability Based Cluster Expansion Oversampling Technique for Imbalanced Data

“…Majority weighted minority oversampling technique (MWMOTE) [9] is effective in selecting hard to learn minority class examples but in this method, small concepts present in minority class examples that are located far from majority class examples are not identified. For handling this problem which is also referred to as within class imbalance in literature, various cluster based methods have been proposed in literature [18][10] [19]. Cluster Based Oversampling (CBO) [19] is an oversampling technique that can handle between-class imbalance and within-class imbalance simultaneously.…”