Fuzzy and smote resampling technique for imbalanced data sets

There are many factors that could affect the performance of a classifier.One of these factors is having imbalanced datasets which could lead to problem in classification accuracy.In binary classification, classifier often ignores instances in minority class.Resampling technique, specifically, unders...

Full description

Bibliographic Details
Main Authors: Zorkeflee, Maisarah, Mohamed Din, Aniza, Ku-Mahamud, Ku Ruhana
Format: Conference or Workshop Item
Language:English
Published: 2015
Subjects:
Online Access:https://repo.uum.edu.my/id/eprint/15646/1/PID160.pdf
Description
Summary:There are many factors that could affect the performance of a classifier.One of these factors is having imbalanced datasets which could lead to problem in classification accuracy.In binary classification, classifier often ignores instances in minority class.Resampling technique, specifically, undersampling and oversampling are the techniques that are commonly used to overcome the problem related to imbalanced data sets. In this study, an integration of undersampling and oversampling techniques is proposed to improve classification accuracy.The proposed technique is an integration between Fuzzy Distance-based Undersampling and SMOTE.The findings from the study indicate that the proposed combination technique is able to produce more balanced datasets to improve the classification accuracy.