The Dependent-variable SMOTE for learning imbalanced data-set

• Main Authors: Yen-ChunChen, 陳彥均
• Other Authors: Der-Chiang Li
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/ccyreb

碩士 === 國立成功大學 === 工業與資訊管理學系 === 106 === Data mining on imbalanced data sets receives more and more attentions in recent years.The class imbalanced problem occurs when there’s just few number of instances in one classes comparing to other classes.The SMOTE:Synthetic Minority Over-Sampling Technique is an effective method to improve the recognition of minority class in class-imbalanced problem. SMOTE is an over-sampling method that generates new synthetic instances from the minority class. And it provides a standard procedure for further research such as Borderline SMOTE , Safe-Level SMOTE and Local-Neighborhood SMOTE . In this paper we propose an extension method of SMOTE by considering the relations between different attributes and deciding the location of synthetic samples based on fuzzy technique.This study develops a new sample-generating procedure to determine the range on the both sides of the line that connects the pairs of population samples.Three data sets taken from UCI Machine Learning Repository in the experiments. We compare the proposed method with SMOTE and other extension version including Borderline SMOTE , Safe-Level SMOTE and Local-Neighborhood SMOTE.The result shows that the proposed method achieves better classifier performance for the minority class than other methods after examined the data sets with C4.5 decision trees.