RACOG and wRACOG: Two Probabilistic Oversampling Techniques

Barnan Das; Narayanan C Krishnan; Diane J Cook

doi:10.1109/TKDE.2014.2324567

Back

RACOG and wRACOG: Two Probabilistic Oversampling Techniques

Journal article

Open access

RACOG and wRACOG: Two Probabilistic Oversampling Techniques

Barnan Das, Narayanan C Krishnan and Diane J Cook

IEEE transactions on knowledge and data engineering, Vol.27(1), pp.222-234

01/01/2015

DOI: https://doi.org/10.1109/TKDE.2014.2324567

Handle:

https://hdl.handle.net/2376/112944

PMCID: PMC4814938

PMID: 27041974

Files and links (1)

url

https://doi.org/10.1109/TKDE.2014.2324567View

Published (Version of record) Open

Abstract

Machine learning algorithms

Approximation algorithms

Probabilistic logic

Probability distribution

Approximation methods

Kernel

Joints

As machine learning techniques mature and are used to tackle complex scientific problems, challenges arise such as the imbalanced class distribution problem, where one of the target class labels is under-represented in comparison with other classes. Existing oversampling approaches for addressing this problem typically do not consider the probability distribution of the minority class while synthetically generating new samples. As a result, the minority class is not represented well which leads to high misclassification error. We introduce two probabilistic oversampling approaches, namely RACOG and wRACOG, to synthetically generating and strategically selecting new minority class samples. The proposed approaches use the joint probability distribution of data attributes and Gibbs sampling to generate new minority class samples. While RACOG selects samples produced by the Gibbs sampler based on a predefined lag, wRACOG selects those samples that have the highest probability of being misclassified by the existing learning model. We validate our approach using nine UCI data sets that were carefully modified to exhibit class imbalance and one new application domain data set with inherent extreme class imbalance. In addition, we compare the classification performance of the proposed methods with three other existing resampling techniques.

Metrics

10 Record Views

See more details

Details

Title: RACOG and wRACOG: Two Probabilistic Oversampling Techniques
Creators: Barnan Das - School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA
Narayanan C Krishnan - Indian Institute of Technology, Ropar, India
Diane J Cook - School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA
Publication Details: IEEE transactions on knowledge and data engineering, Vol.27(1), pp.222-234
Academic Unit: Electrical Engineering and Computer Science, School of
Publisher: IEEE
Grant note: 1064628 / National Science Foundation (10.13039/100000001) R01EB009675 / National Institutes of Health (10.13039/100000002)
Identifiers: 99900547674301842
Language: English
Resource Type: Journal article