症状监测预警数据分析及方法研究

发布时间：2018-01-13 20:40

本文关键词：症状监测预警数据分析及方法研究　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：症状监测对疾病预防具有日益显著的前置性优势,能够在大规模疾病爆发之前感知到异常趋势,为医疗卫生预防工作留有更加充足的准备时间。近年来,症状监测作为一种全新的监测手段,正越来越多地引起人们的关注。但是现有的疾病监测方式有严重的滞后性,在信息录入形式上的便捷性及信息化程度低,且采集到的样本信息没有准确的数据处理方式及预警分析方法。因此,为了建立可行的症状监测预警体系,实现对重点区域的实时监测和数据分析,本文应用准确的预警模型及大数据关联信息挖掘方法对监测数据进行数据分析、处理和预警响应,得到快速、准确的预警结果。论文由几个部分构成:(1)采用专家评价法筛选设计了症状监测系统的数据源,并运用脏数据清洗的方式对采集数据进行粗略的预处理。(2)从似然比的角度设计了休哈特控制图,基于其对数据中小偏移监测不灵敏的特点提出了累积和控制图,并完成了监控方差的CUSUM控制图模型设计,同时提出用马尔可夫链的方式计算控制图参数组合下的ATS和ARL值,用以对比模型的灵敏度。(3)设计了改进型的具有变动抽样区间的动态CUSUM控制图,通过ATS及ANSS值的比较得出动态控制图具有更好的检出效果,并提出了差异率的概念,对比后得出不同参数组合下动态控制图都具有优于静态控制图的预警效果。(4)将监控方差的变动抽样区间的CUSUM控制图用于症状监测采样数据,使用灰色关联为症状数据加权分析,并人为设定异常点,改变采样数据的方差,得出动态控制图具有准确度高、灵敏度高的检出效果。论文采用马氏链的方法计算相应的ATS及差异率,得出采用长短抽样区间进行控制图设计具有更好监测效果的结论,并将其应用于症状监测预警系统之中,有效的提高了报警灵敏度,降低了报警时间,提高了监测效率。
[Abstract]:Symptom monitoring has an increasingly significant predominance in disease prevention, and it can perceive abnormal trends before large-scale disease outbreaks, leaving more adequate preparation time for medical and health prevention work in recent years. As a new monitoring method, symptom monitoring is attracting more and more attention. However, the existing disease surveillance methods have serious lag, and the convenience and information level of information entry is low. And the collected sample information has no accurate data processing and early warning analysis method. Therefore, in order to establish a feasible symptom monitoring and early warning system, real-time monitoring and data analysis of key areas can be realized. In this paper, the accurate early warning model and big data correlation information mining method are applied to the monitoring data analysis, processing and early warning response, and get fast. The paper is composed of several parts. (1) the data source of the symptom monitoring system is designed by the expert evaluation method. And using dirty data cleaning method to collect data rough preprocessing. 2) from the angle of likelihood ratio design the control chart of Shewhart. Based on its insensitivity to small and medium offset monitoring, the cumulative and control charts are proposed, and the CUSUM control chart model of monitoring variance is designed. At the same time, the method of Markov chain is used to calculate the values of ATS and ARL under the control chart parameter combination. A modified dynamic CUSUM control chart with variable sampling interval is designed to compare the sensitivity of the model. Through the comparison of ATS and ANSS, it is concluded that the dynamic control chart has better detection effect, and the concept of difference rate is put forward. After comparison, it is concluded that the dynamic control chart with different parameters combination has better warning effect than static control chart.) the CUSUM control chart of the variable sampling interval of the monitoring variance is applied to the symptom monitoring sampling data. Grey correlation is used as the weighted analysis of symptom data, and the abnormal points are set artificially to change the variance of the sampling data, and the dynamic control chart is obtained with high accuracy. The method of Markov chain is used to calculate the corresponding ATS and the difference rate. It is concluded that the control chart design with long and short sampling interval has better monitoring effect. It is applied to the symptom monitoring and warning system, which effectively improves the alarm sensitivity, reduces the alarm time and improves the monitoring efficiency.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R181.8;TP277

【参考文献】