一种基于密度的SMOTE方法研究.PDFVIP

第 12 卷第 6 期 智　能　系　统　学　报 Vol.12 No.6 2017 年 12 月 CAAI Transactions on Intelligent Systems Dec. 2017 DOI: 10.11992/tis.201706049 网络出版地址: /kcms/detail/23.1538.TP1534.030.html 一种基于密度的SMOTE 方法研究 王俊红，段冰倩 （山西大学 计算机与信息技术学院，山西 太原 030006 ） 摘 要：重采样技术在解决非平衡类分类问题上得到了广泛的应用。其中，Chawla 提出的SMOTE(Synthetic Mino- rity Oversampling Technique) 算法在一定程度上缓解了数据的不平衡程度，但这种方法对少数类数据不加区分地进 行过抽样，容易造成过拟合。针对此问题，本文提出了一种新的过采样方法：DS-SMOTE 方法。DS-SMOTE 算法基 于样本的密度来识别稀疏样本，并将其作为采样过程中的种子样本；然后在采样过程中采用SMOTE 算法的思想，在 种子样本与其k近邻之间产生合成样本。实验结果显示，DS-SMOTE 算法与其他同类方法相比，准确率以及G值有 较大的提高，说明DS-SMOTE 算法在处理非平衡数据分类问题上具有一定优势。 关键词：非平衡；分类；采样；准确率；密度 中图分类号：TP311 文献标志码：A 文章编号：1673−4785(2017)06−0865−08 中文引用格式：王俊红, 段冰倩. 一种基于密度的SMOTE 方法研究[J]. 智能系统学报, 2017, 12(6): 865–872. 英文引用格式：WANG Junhong, DUAN Bingqian. Research on the SMOTE method based on density[J]. CAAI transactions on in- telligent systems, 2017, 12(6): 865–872. Research on the SMOTE method based on density WANG Junhong，DUAN Bingqian (School of Computer and Information Technology, Shanxi University, Taiyuan 030006, China) Abstract : In recent years, over-sampling has been widely used in the field of classification of imbalanced classes. The SMOTE(Synthetic Minority Oversampling Technique) algorithm, presented by Chawla, alleviates the degree of data im- balance to a certain extent, but can lead to over-fitting. To solve this problem, this paper presents a new sampling meth- od, DS-SMOTE, which identifies sparse samples based on their density and uses them as seed samples in the process of sampling. The SMOTE algorithm is then adopted, and a synthetic sample is generated between the seed sample and its k neighbor. The proposed a