基于加权模糊聚类的不平衡数据分类方法  

An Imbalanced Data Classification Method Based on Weighted Fuzzy Clustering

作　　者：董振波[1] 

机构地区：[1]重庆大学计算机学院,重庆400030

出　　处：《现代计算机（中旬刊）》2016年第6期25-27,共3页Modern Computer

摘　　要：不平衡数据分类问题是数据分类的一个热门话题,经常出现在数据分类问题的实践当中,而且该问题给传统的分类方法提出了巨大的挑战。由于在分类过程中,样本对正负类的归属往往比较模糊,提出一种基于模糊聚类的不平衡数据分类方法,同时为了增强对正负类的区分能力,将基于类差异的属性加权方法引入到该方法中。该方法在通过模糊C均值算法得出样本对正负类的归属程度后根据测试样本对正负类的相似性对其进行分类。Imbalanced data sets classification is a hot topic in data mining field. It often appears in the practice of data classification, and it is a great challenge for traditional classification methods. Because during classification processes, which class the samples are belonged to, is not always so clear, so proposes a fuzzy clustering based classification method for imbalanced data. At the same time, in order to enhance the ability to distinguish positive and negative class, adopts a feature weighting method based on class dissimilarity. This method uses the fuzzy C-mean algorithm to obtain the degrees the samples belonging to each class and then decides which class the training samples be-long to.

关 键 词：不平衡数据集分类 模糊集 Hellinger距离 

分 类 号：O159[理学—数学]