基于统计学习理论的传染病预警方法研究比较

发布时间：2018-01-18 06:09

本文关键词：基于统计学习理论的传染病预警方法研究比较　出处：《辽宁师范大学》2016年硕士论文　论文类型：学位论文

【摘要】：在所有的突发公共卫生安全事件中,影响最严重的当属传染病疫情的爆发。传染病疫情的爆发不仅影响人们的正常出行,还容易造成社会大众的心理恐慌,造成国家经济和社会方方面面的动荡局面。而传染病预测预警技术的产生与发展,对防范传染病爆发,及时有效的采取应对措施就显得尤为重要。在诸多统计学学习理论中,应用于预测预警方面的方法有很多并且表现都不错。本文主要选取了三种常用的统计模型,选取传染病中的结核病为对象,通过对收集到的结核病数据进行统计分析,应用10折交叉检验的方发比较得出在结核病预测预警方面表现比较好的模型。本文首先介绍了突发传染病事件近些年来的状况以及它给人们的生活带来的影响。同时介绍了各个国家面对突发事件做出了哪些应对措施,通过分析各个国家应用不同统计方法预测传染病模型的优缺点,最终本文选择了三种统计模型,以结核病为传染病的代表,比较分析这三种模型在传染病的预测预警方面的表现。其次,分别介绍了结构方程模型、人工神经网络模型和随机森林模型这三种模型的基础理论知识。最后进行实证分析。应用R软件程序语言,对收集到的大连市各个学校的结核病数据进行数据处理分析。结果得到结构方程模型可以将不可观测的隐变量用多个可观测的指标表示出来,而且能很好的表示隐变量之间的因果关系。但结构方程的局限性是它本身是一种验证性的模型,因此不同的模型可能出现不同的结果。人工神经网络的非线性处理能力很好,但是容易出现过拟合现象。随机森林处理数据过程快,而且不容易发生过拟合现象。从人工神经网络和随机森立的10折交叉验证结果显示,随机森林模型的拟合度更高,更稳定。
[Abstract]:In all the sudden public health and safety incidents, the most serious impact is the outbreak of infectious disease. The outbreak of infectious disease not only affects the normal travel of people, but also easily cause psychological panic of the general public. The emergence and development of early warning technology for infectious diseases will prevent the outbreak of infectious diseases. Timely and effective measures are particularly important. In many statistical learning theory. There are many methods used in forecasting and early warning, and the performance is good. This paper mainly selects three commonly used statistical models, select tuberculosis in infectious diseases as the object. Through the collection of tuberculosis data for statistical analysis. By using the 10% cross test, the model of TB prediction and early warning is obtained. This paper first introduces the situation of emergent infectious diseases in recent years and its impact on people's lives. At the same time, it introduces the response measures that each country has made in the face of emergency. By analyzing the advantages and disadvantages of using different statistical methods to predict infectious disease models in different countries, three statistical models were selected in this paper, with tuberculosis as the representative of infectious diseases. The performance of these three models in the prediction and early warning of infectious diseases is compared and analyzed. Secondly, the structural equation models are introduced respectively. The basic theoretical knowledge of artificial neural network model and stochastic forest model. Finally, the empirical analysis. Using R software programming language. The data of tuberculosis collected from every school in Dalian were analyzed. The results showed that the unobservable hidden variables could be expressed by multiple observable indexes in the structural equation model. But the limitation of structural equation is that it is a kind of confirmatory model. Therefore, different models may have different results. Artificial neural network has good nonlinear processing ability, but it is prone to over-fit phenomenon. The data process of stochastic forest processing is fast. From the results of artificial neural network and 10 fold cross validation of random forest model, the fitting degree of stochastic forest model is higher and more stable.
【学位授予单位】：辽宁师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R181.8

【相似文献】