基于高斯混合模型的心血管患者虚弱程度度量模型  

Measurement model of frailty degree of cardiovascular patients based on Gaussian mixture model

作　　者：冯云霞[1] 韩正亮 刘少阳 宋波[1] FENG Yunxia;HAN Zhengliang;LIU Shaoyang;SONG Bo(College of Information Science and Technology,Qingdao University of Science&Technology,Qingdao Shandong 266061,China)

机构地区：[1]青岛科技大学信息科学技术学院,山东青岛266061

出　　处：《计算机应用》2020年第S02期230-236,共7页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61572268,61303193);山东省重点研发计划项目(2017GSF18110,2018GGX101029)。

摘　　要：针对度量心血管疾病致虚弱症效率低、准确率不高的问题,提出了一种基于概率密度分析的心血管患者虚弱程度度量模型。首先,通过核主成分分析对患者数据的特征矩阵进行升维,提取其线性相关性;然后,利用主成分分析对升维后的矩阵进行降维,得到度量模型的输入指标;最后,利用高斯混合模型初步度量患者的虚弱程度,对其中决策边界附近的数据再次利用高斯混合模型提升度量精度。利用心血管疾病致虚弱症患者数据检验度量模型的应用效果。实验结果表明,所提模型的误差率约为2.9%,正确率较虚弱量表提高了5.6个百分点,同时运行效率提升了13个百分点。该模型能够有效度量心血管患者的虚弱程度,同时该模型具备一定的可扩展性,可实现更细致的程度划分。Focusing on the issue that cardiovascular disease-induced frailty measurement is inefficient and inaccurate,a new model for measuring the frailty degree of cardiovascular patients based on probability density analysis was proposed.Firstly,the characteristic matrix dimension of patient data was increased by Kernel Principal Component Analysis(KPCA)and its linear correlation was obtained.Then,Principal Component Analysis(PCA)was used to reduce the dimensionality of matrix and the input metrics of measurement model were got.Finally,Gaussian mixture model was used to measure the frailty degrees of patients.For the data neared the decision boundary,Gaussian mixture model was used again to improve the measurement accuracy.The application effect of the measurement model was tested by using the data of patients with cardiovascular disease-induced frailty.The simulation results show that,the error rate of the proposed model is about 2.9%,its accuracy is increased by 5.6 percentage points compared with the frailty scale,and the operation efficiency is increased by 13 percentage points.The application of the proposed model for measuring frailty degree caused by cardiovascular disease has a better effect.And the proposed model has good scalability,and can achieve a more refined classification of frailty.

关 键 词：心血管疾病 虚弱症 核主成分分析 高斯混合模型 无监督学习 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]