BAM: BORDER ADJUSTMENT METHOD IMPROVE THE EFFICIENCY OF IMBALANCED BIOLOGICAL DATA CLASSIFICATION

Tóm tắt

This paper presents a data classification problem and methods to improve imbalanced data classification. Especially, biomedical data has a very high imbalance rate and the sample identification of minority class is a very important. Many studies have shown that border elements are important in imbalanced data classification such as Borderline-SMOTE, Random Under Border Sampling. This paper provides a new method of adjusting data: generating synthetic elements on the borderline of the minority class, identify and eliminate noise elements of the majority class to achieve better classification efficiency. Experimental results of classification of SVM algorithm on six datasets of UCI international standard data warehouse: Blood, Haberman, Pima, Yeast, Ionosphere, and Glass showed that the adjustment of borderline has a positive effect on classification and the results are considered statistically significant.