Abstract
The class imbalance problem corresponds to the problem encountered by inductive learning systems on domains for which one class is represented by a large number of examples while the other is represented by only a few. When learning from such skewed/imbalanced data, traditional methods tend to classify all data into the majority class, which is usually the less important class. In machine learning, this problem is quickly becoming an emerging issue of practical areas, such as fraud detection, failure monitoring/inspection, text mining, and medical diagnosis. This study proposes two kinds of Granular Computing approaches to tackle this problem. A real case of diabetes diagnosis is provided to illustrate the effectiveness of the proposed methods. Compared with traditional methods, Support Vector Machines (SVM) and under-sampling method, experimental results show our approach can remarkably improve the ability of detecting minority examples without losing overall classification accuracy.