【作者】 艾书超；

【导师】 刘舒燕；

【作者基本信息】 武汉理工大学 ， 系统工程， 2005， 硕士

【摘要】 本论文主要讨论用于样本外预测的时间序列模型的构建。时间序列是对经济变量的观测,而这些变量可能源于各个不同的商业与经济领域。 时间序列数据多式多样。典型的情况是,许多宏观的经济总量指标,显示出一种向上的趋势;而旅游业数据和零售业数据则呈现出显著的季节性。时间序列的另一个特点是经济数据的某些观察值在某种意义上来说可能是异常数据。 显然,不存在一个简单的时间序列模型,能够描述上面所有特征,而且还能够获得相当精确的样本外预测值。本论文的主要意图仅仅只是针对经济学领域的某个具体的问题,讨论如何建立相应的时间序列模型以及怎样评价那些模型在预测中的优点。 本论文的结构: 第一章 绪论 第一节简述本论文选题的目的、内容。第二节阐述了时间序列的概念简述、分析方法的分类,以及研究现状。第三节是本论文研究的目标、内容和准备采取的方法。 第二章 时间序列数据介绍 第一节阐述时间序列数据的一般特征:趋势性、周期性、异常观测值、条件异方差、非线性、共同的趋势等等;第二节是时间序列分析采用的统计学方法分类;第三节重点介绍时间序列分析的相关基本概念。 第三章 时间序列的趋势分析 第一节论述确定性模型(DT)与随机趋势模型(ST)的定义以及建模方法;第二节在讨论单位根的检验;第三节讨论平稳性检验,以此来决定是选择DT还是ST模型; 第四章 异常观测值处理 在本章中,我们论述了几种时间序列异常值,即可加的异常值(AO)、新生的异常值(IO)、永久的水平迁移(PLS)、趋势改变(CT)等,并叙述了这几种模型各自相应的建模方法。 第五章 实证分析 本章中,我们采用时间序列分析法来研究我国高速公路发展。我们从源数据入手,结合前面几章的理论,对时间序列采取多种数据处理手段,缩小模型选择范围,逐步建立了我们认为比较适合的时间序列模型。我们又对模型进行了一定的修正,并进行了异常值分析,结果更加准确。更多还原

【Abstract】 This paper would be used to research the methods for constructing the models of time series to forecast out of the samples. The time series are surveying value of economic varies, and this varies maybe collected from various fields. There are many methods of time series’ researching. In this paper, we only use the methods in measurement economics.There are a lot of features for time series. As typical cases, some macroeconomic amount indexes; and the data of tour industry and retailing industry show the marked seasonality. Another characteristic is that observation of economic data maybe looks as if an abnormal value in some significance.Clearly, this kind of models that is enough simple to describe all the features above said, but to predict the value out of samples extremely accurate. The construction of this paper: The first chapter: prefaceWe describe our purpose of this paper in the first section. In the second section, we introduce the conception of the time series, classify of researching methods, the present conditions in international. In third section, we told our objects, content, and the methods we would take. The second chapter: some introductions of time seriesIn first section, the main features of time series would be described. In second section, classify of the methods for analyzing the time series list. Then, we present some definitions of some conceptions on time series. The third chapter: trend analysis of time seriesIn this chapter, we talk about how to analyze the trend of a time series, how to construct the ST model or DT model, and compare two kind of the models. The forth chapter: Abnormal value analysisWe discus some kind of abnormal value of time series, that is additive outlier (AO), innovative outlier (IO), permanent level shift (PLS), changing trend (CT).Then we discus how to construct these models. The fifth chapter: the case of time series analysisWe present the data of the highway development of our country. On the basis of the theory, we use some ways, that is natural log transformation and difference transformation, we construct a exact model step by step. Lastly, we modify the model and make the forecasting.更多还原