安徽建筑大学
时间序列分析课程设计报告书
院系数理学院
专业统计学
班级统计学三班
学号11207040302
姓名朱敏
指导教师俞泽鹏
基于时间序列分析的股票预测模型研究
摘要
在现代金融浪潮的推动下,越来越多的人加入到股市,进行投资行为,以期得到丰厚的回报,这极大促进了股票市场的繁荣。而在这种投资行为的背后,越来越多的投资者逐渐意识到股市预测的重要性。所谓股票预测是指:根据股票现在行情的发展情况地对未来股市发展方向以及涨跌程度的预测行为。这种预测行为只是基于假定的因素为既定的前提条件为基础的。但是在股票市场中,行情的变化与国家的宏观经济发展、法律法规的制定、公司的运营、股民的信心等等都有关联,因此所谓的预测难于准确预计。即使是证券分析师的预测也只能作为股民入市操作的一般参考意见。时间序列数据因为接受到许多偶然因素的影响,会常常表现出随机性,在统计学上称之为序列的依赖关系。时间序列分析是经济预测领域研究的重要工具之一,它描述历史数据随时间变化的规律,并用于预测经济数据。在股票市场上,时间序列预测法常用于对股票价格趋势进行预测,为投资者和股票市场管理管理方提供决策依据。本文主要介绍了时间序列分析方法的概念,性质,特点以及时间序列模型,包括建模时对数据时间序列的预处理、模型识别、参数估计、模型检验、模型优化以及模型预测等。并根据道琼斯指数对收盘价进行短期预测,通过对时间序列分析理论的实证研究分析,建立时间序列模型,说明时间序列分析的方法对于股票价格
的预测趋势有一定的参考价值。
关键词：股票，预测，时间序列分析，AR（1 ）模型
ABSTRACT
In the modern financial wave, more and more people join the stock market to invest, expecting to get rich return, which has greatly promoted the stock market’s prosperity. While under this behavior, an increasing large number of people become to realize the importance of stock forecast. The so-called stock forecast is defined: with the help of the stock’s recent condition, we’ll predict the future stock’s development, including its later development directions and fluctuations. This prediction based on the assumption of behavior is the prerequisite for established factor basis. But the stock’s index is always changing with the country’s macroeconomic development, the formulation of laws and regulations, the company’s operations, the confidence of investors and so on, which results in that it is very difficult to accurately predict. Even securities analysts’forecast results can only be operated as a general reference. Time-series data often show some kinds of randomness and dependence between each other because of the influence of various accidental factors. Time series analysis is one of the most important tools for economy research, and it describe the variation of data with time, and used to forecast economic data.Time series analysis is often used to predict the stock price, which provides decision-making basis for investors and the stock market managers. This thesis mainly introduces time series analysis theory, including its notion, character as well as the expression and description of some models derived from it ,including method of data simulation, method of parameter estimation and method of testing degree of fitting and arrange them by the numbers. And according to the Dow Jones
index, we may predict the closing price trend for short-term with the help of time series analysis theory. Therefore we can establish some models, we could prove that the method has some value for predicting the stock’s trend by means of model fitting effect and error analysis.
Keywords: stock, predict, time series analysis, AR（1）model
