多源健康数据的语义分析方法研究

发布时间：2018-05-29 12:52

本文选题：多源数据融合 + 感知数据　；参考：《西北工业大学》2016年博士论文

【摘要】：网络技术的诞生将人类在物理空间的交互拓展延伸到了虚拟的信息空间中,克服了空间对于人类的制约作用,大大缩短了信息传播时间,强化了用户之间的交互。随着感知技术的进一步发展和移动计算终端的出现,人类活动在两个空间中不断地交替切换,从而使得物理空间和信息空间的互动更加频繁,并在两个空间中留下了大量的数字信息。多个空间的交融促进了Cyber-Physical-Social System(CPSS)的出现。CPSS的多源数据空间具有异构性、动态性、稀疏性和高噪声等特性,其对数据表示、处理和分析等提出了诸多挑战。本文以健康数据分析为应用背景,研究了从物理空间中的感知数据和信息空间中的在线数据中抽取与健康相关的语义信息方法。本文主要研究工作与创新如下:多源数据及其计算模型多源数据空间包含多种多样的数据信息,这些数据具有异构性、动态性、稀疏性、高噪声等特点。本文分析了感知数据和在线数据的特点,分别定义了感知数据模型和在线数据模型,实现不同特质数据的表示和存储。为了实现不同类型数据的混合计算,本文采用本体表示模型抽象不同数据之间的关系,并基于知识推理构建相应的计算模型。轻量级的感知数据健康语义分析随着大量生理感知器件的使用,利用感知设备对用户健康状态进行检测和分析成为可能。本文重点研究了基于移动设备监控用户运动水平的方法。首先,本文对用户行为识别研究从感知技术和识别算法的角度进行了综述性的分析。重点研究轻量级的行为分析方法,以达到长期监控运动状态的目的。本文从采样频率、特征选择和算法角度入手,针对移动终端资源受限的特点,提出了一个轻量级层次化的行为分析算法,其通过降低采样频率和频域数据的使用概率,实现识别正确率和能耗之间的平衡。基于该行为分析算法,本文使用代谢当量分析方法实现用户的运动状态评估。基于词向量的在线数据健康语义分析用户在线活动是人类物理交互在网络空间的延伸,其中可能隐含用户的健康信息。本文研究了从大量的在线数据尤其是文本数据中挖掘用户健康行为的方法。首先研究了在线数据采集的策略和方法,构建了一个社交媒体数据采集平台,实现从不同的社交服务平台采集数据信息。然后,从交互特征、语义特征和拓扑特征等角度分析了在线数据与用户健康行为属性之间的关系。最后借助词向量的文本表示技术,以控烟为应用背景,提出了由在线数据判断用户是否为烟草产品使用者的算法。层次化的多源健康数据融合方法针对感知数据和在线数据的差异,本文研究了CPSS系统中的多源数据融合问题,从特征级融合和决策级融合两个层次分析了多源数据融合方法。针对决策级融合,本文采用知识推理方法抽取高级语义信息;针对特征融合,本文基于多源特征提出了一个社群发现算法,旨在找到具有强交互或者共同话题偏好的社群。面向老年人的多源数据融合应用验证结合我国人口老龄化的严峻状况,本文设计实现了一个基于多源数据融合的社会化用药提醒应用原型系统。该系统借助于多源数据可以感知用户的运动和健康状态,能够检测用户的用药情况,依据用户交互关系和共同偏好选择合适的提醒发起者,实现自适应的用药提醒。本文从系统架构的角度对系统中各个层次的功能需求进行阐述,并对原型系统进行初步的分析和评估。
[Abstract]:The birth of network technology extends the human interaction extension in the physical space into the virtual information space, overcomes the restriction of space to the human, greatly shortens the time of information transmission and strengthens the interaction between users. With the further development of the perceptual technology and the emergence of the mobile computing terminal, the human activities are in two spaces. Alternately alternately, the interaction between physical space and information space is more frequent, and a large number of digital information is left in the two spaces. The integration of multiple spaces promotes the appearance of Cyber-Physical-Social System (CPSS) in the multisource data space of.CPSS, which has the characteristics of heterogeneity, dynamics, sparsity and high noise, and so on. A number of challenges are presented for data representation, processing and analysis. This paper, taking health data analysis as the application background, studies the extraction of health related semantic information from the perceptual data in the physical space and the online data in the information space. The main research work and creation are as follows: multisource data and the multisource number of its computing models. Space contains a variety of data information. These data are heterogeneous, dynamic, sparse, and high noise. This paper analyzes the characteristics of perceptual data and online data, defines the perceptual data model and the online data model, and realizes the representation and storage of different trait data. In order to realize the mixing of different types of data In this paper, the relationship between different data is abstracted from the ontology representation model and the corresponding calculation model is constructed based on knowledge reasoning. The lightweight perceptual data semantic analysis is possible with the use of a large number of physiological sensing devices, and it is possible to detect and analyze the health status of the users by using the perceptual devices. This paper focuses on the research of the base. First, this paper makes a summary analysis of user behavior recognition from the perspective of perceptual technology and recognition algorithm. This paper focuses on lightweight behavioral analysis methods to achieve long-term monitoring of motion status. This paper starts with sampling frequency, feature selection and algorithm angle. A lightweight hierarchical behavior analysis algorithm is proposed for the characteristics of mobile terminal resource constraints. By reducing the probability of using the sampling frequency and frequency domain data, the balance between the recognition accuracy and the energy consumption is realized. Based on the behavior analysis algorithm, this paper uses the metabolic equivalent analysis method to realize the user's motion state evaluation. The online data health semantic analysis based on word vector is an extension of human physical interaction in the network space, which may imply the user's health information. This paper studies the method of mining user health behavior from a large number of online data, especially text data. The first research on the strategies and parties of online data acquisition is made. A social media data collection platform is constructed to collect data from different social service platforms. Then, the relationship between the online data and the user's health behavior attributes is analyzed from the aspects of interactive features, semantic features and topological features. In this paper, the multisource data fusion problem in CPSS system is studied in this paper. The multisource data fusion method is analyzed from two levels of feature level fusion and decision level fusion. Based on the feature fusion, this paper proposes a community discovery algorithm based on multi source feature, aiming at finding a community with strong interaction or common topic preference. The application verification of multi-source data integration for the elderly combines the grim shape of the aging population in China. In this paper, a social drug reminder application prototype system based on multi source data fusion is designed and implemented. The system can perceive the user's movement and health state with the aid of multi source data, can detect the user's medication situation, and select the appropriate reminder based on the user interaction and common preference, and realize the adaptive drug use. Reminding. From the point of view of system architecture, this paper expounds the functional requirements of all levels in the system, and makes preliminary analysis and evaluation of the prototype system.
【学位授予单位】：西北工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.1

【相似文献】