基于ICA改进ICEEMD的UDS重采样数学模型

A Mathematical Model for Resampling Unbalanc ed Data Set Based on ICA and Improved ICEEMD

为了增强不平衡数据集处理效果,提出一种基于ICA改进ICEEMD的不平衡数据集重采样数学模型研究方法。分析不平衡数据集的分布特征,通过改进完备集成经验模态分解(ICEEMD)方法和独立分量分析(ICA)分解不平衡数据集,去除不平衡数据集中的噪声。通过DP聚类算法和σ准则构建重采样数学模型,利用该模型自动判别不平衡数据集的聚类中心和离群点,同时对多数和少数类样本分析处理,确保样本相对均衡,最终完成不平衡数据集的重采样处理。经实验测试结果表明,所提模型的整体性能明显优于其它重采样模型,验证了其应用价值。

In order to improve the effect of processing the unbalanced data sets,this paper put forward a method of researching the resampling mathematical model of unbalanced data sets based on improved ICEEMD-ICA.Firstly,the distribution characteristics of the unbalan ced data set was analyzed.And then,the Impro ved Complementary Ensemble Empirical Mode Decomposition(ICEEMD)and In dependent Component Analysis(ICA)were used to decompose the unbalanced data set and thus to remove the no ise from it.Secondly,DP clustering algorithm andσcriterion were adopted to construct a resampling mathematica l model,which could automatically identify t he cluster center and outliers of the unbalanced data set and analy ze the majority and minority samples at the s ame time,thus ensuring that the samples were relatively balanced.Finall y,the resampling process for the unbalanced data set was completed.The experimental results show that the overall p erformance of the proposed model is signific antly better than other models.