面向时间序列相似性的疾病风险评估模型  

Disease Risk Assessment Model Based on Similarity Measurement of Multivariate Time Series

作　　者：沈彧 陈庆奎[1] SHEN Yu;CHEN Qing-kui(School of Optical-Electrical Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《小型微型计算机系统》2022年第9期1869-1876,共8页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61572325)资助;上海重点科技项目(19DZ1208903,2018RGZN02061)资助;上海智能家居大规模物联共性技术工程中心项目(GCZX14014)资助.

摘　　要：为进一步提升养老机构的服务水平,提出一种利用多元时间序列相似性度量对老人疾病风险进行评估的模型:首先通过离散小波变换将老人监测序列分解为尺度部分和细节部分,然后在完成对序列分段的基础上,提取局部极值特征对序列进行降维,同时根据老人行为存在规律的特点,运用自适应惩罚函数来减少单点的重复使用以及匹配点间相位差过大的情况,以最大程度地识别不同老人相似的行为;再通过主成分分析确定不同检测属性的权值,得到多元序列的相似性距离;最后结合提出的相似性度量对老人的状态检测序列进行K近邻分析,利用分类的结果完成对老人的疾病风险评估.实验结果表明,该方法的分类准确率达到91%以上,能对老人的疾病风险进行合理评估.In order to further improve the service level of elderly care institutions,this paper proposes a disease risk assessment model for the elderly based on multiple time series similarity measures:First,the discrete wavelet transform is used to decompose the elderly monitoring sequence into the scale part and the detail part.On the basis of completing the segmentation of the sequence,the local extreme value features are extracted to reduce the dimensionality of the sequence.At the same time,according to the characteristics of the regular behavior of the elderly,the adaptive penalty function is used to reduce the repeated use of single points and the situation where the phase difference between matching points is too large,to maximize the identification of similar behaviors of different elderly;Then the weights of different attributes are determined by principal component analysis,and the similarity distance of the multivariate sequence is obtained;Finally,combined with the proposed similarity measure,K-nearest neighbor analysis is performed on the elderly′s staged behavior,and the classification results are used to complete the disease risk assessment of the elderly.Experimental results show that the classification accuracy of this method is more than 91%,and it can reasonably evaluate the disease risk of the elderly.

关 键 词：养老服务 相似性度量 多元时间序列 风险评估 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]