一种适用于高维非线性特征数据的聚类算法及应用

发布时间：2018-07-20 17:14

【摘要】：针对高维数据聚类分析中数据之间具有多种非线性特征关系,导致数据分布不均、传统相似性度量失效及结果类中心难以精准表征等问题,提出了一种基于核主元分析(KPCA)与密度聚类(DBSCAN)的高维非线性特征数据聚类分析技术。首先,为有效提取高维数据的非线性特征,利用KPCA理论将原始数据映射到更高维数据空间,利用主元分析获得数据变化的方向集合,并进行降维分析;然后,通过重新定义数据样本在主元空间的相似性距离对传统DBSCAN聚类方法进行改进,并利用3δ统计理论对各簇中心的进行表征,从而实现高维数据的精确分类与类中心知识表达。以实际高血压患者群体聚类问题为例对方法进行了有效性验证,实验表明,所提方法可以有效获取原始数据的非线性特征,实现患者个体特征群体的有效划分及簇类中心知识的表达,解决传统DBSCAN聚类方法对高维数据不适用的问题。
[Abstract]:In high dimensional data clustering analysis, there are many nonlinear feature relationships between the data, which lead to uneven distribution of data, failure of traditional similarity measurement and difficulty in accurate representation of result class centers, and so on. A high dimensional nonlinear feature data clustering technique based on kernel principal component analysis (KPCA) and density clustering (DBSCAN) is proposed. Firstly, in order to extract the nonlinear features of high-dimensional data effectively, the original data is mapped to higher-dimensional data space by using KPCA theory, and the direction set of data change is obtained by principal component analysis, and then the dimensionality reduction analysis is carried out. The traditional DBSCAN clustering method is improved by redefining the similarity distance of data samples in the principal component space, and each cluster center is represented by using the 3 未 statistical theory to realize the accurate classification of high-dimensional data and the knowledge representation of the cluster center. The validity of the method is verified by taking the cluster problem of real hypertension patients as an example. The experimental results show that the proposed method can effectively obtain the nonlinear characteristics of the original data. In order to solve the problem that traditional DBSCAN clustering method is not applicable to high dimensional data, the effective division of patient individual characteristic groups and the expression of cluster center knowledge are realized.
【作者单位】：西安交通大学机械制造系统工程国家重点实验室;西安交通大学第二附属医院;
【基金】：国家自然科学基金资助项目(51375375) 中央高校基本科研业务费专项资金资助项目(xjj2014108)
【分类号】：TP311.13

【相似文献】