基于稀疏子空间聚类算法的高维数据聚类研究  

High-dimensional data clustering based on sparse subspace clustering algorithm

作　　者：王成礼 王洁 陈乃健 WANG Chengli;WNAG Jie;CHEN Naijian(Shenzhen Pingshan District Central Hospital,Pingshan District,Shenzhen,Guangdong 518118,China;Zhongshan Boai Hospital.Zhongshan City,Zhongshan,Guangdong 528403,China)

机构地区：[1]深圳市坪山区中心医院,深圳518118 [2]中山市博爱医院,广东528403

出　　处：《自动化与仪器仪表》2025年第1期84-88,共5页Automation & Instrumentation

基　　金：广东省中山市重大科技计划项目(2018A2KC115)。

摘　　要：针对医疗数据规模大、维度高的问题,由于采用传统的聚类算法对其处理计算复杂度较高,且准确率较低。研究基于稀疏子空间聚类算法设计了一种医疗数据分类方法,并引入了无监督度量学习对分类中的预处理过程进行优化,提出一种结合稀疏子空间聚类算法和无监督度量学习的高维医疗数据分类方法。结果显示,设计方法的平均概率兰德指数为0.85,高于其他算法,设计方法的平均信息变化指数为1.54,低于其他算法,证明其鲁棒性较强。在不同数据集上,设计方法的误分率分别为1.2%和0.9%,证明了其分类精度较高。设计方法在处理高维医疗数据方面具有较高的可靠性,其能够在医疗数据分析领域发挥重要作用,并为精准医疗、疾病预测和诊断提供有力的支持。In response to the problem of large scale and high dimensionality of medical data,traditional clustering algorithms have high computational complexity and low accuracy in processing it.A medical data classification method was designed based on sparse subspace clustering algorithm,and unsupervised metric learning was introduced to optimize the preprocessing process in classification.A high-dimensional medical data classification method combining sparse subspace clustering algorithm and unsupervised metric learning was proposed.The results show that the average probability Rand index of the design method is 0.85,higher than other algorithms,and the average information change index of the design method is 1.54,lower than other algorithms,indicating its strong robustness.On different datasets,the misclassification rates of the design method were 1.2%and 0.9%,respectively,proving its high classification accuracy.The design method has high reliability in processing high-dimensional medical data,and can play an important role in the field of medical data analysis,providing strong support for precision medicine,disease prediction,and diagnosis.

关 键 词：医疗数据 高维 稀疏子空间聚类 无监督度量学习 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]