时间序列分析中数据的平稳性判定研究

摘要

时同序列的平稳性判定是时间序列分析预测的关键技术,为了根据数据特征提供更为可靠合理的平稳性判定方法,从数据平稳条件入手比较分析了时间路径图、自相关函数、DF检测和ADF检测四种方法的数学原理.以股票数据为应用背景,采用EViews工具对时间序列的平稳性判定进行了实验仿真和对比分析,得出对于复杂的时间序列多种检测方法综合检验更为可靠的结论,为随机过程中数据分析预测的进一步研究提供数据预处理的技术参考..%The stability identification of time sequence is a key technology of time sequence analysis and forecasting. To provide more reliable stability judging method according to data feature, the mathematic principles of time path graph, autocorre-lation function, DF test and augment dickey-fuller test(ADF) test are analyzed proceeding from the conditions of data stability. Taking the stock data as the application background, simulation experiment and contrastive analysis for time sequence stability decision are performed with EViews tool. A conclusion that the comprehensive test including multiple detection methods is more reliable for the complex time sequence was obtained. The research provided a data preprocessing technical reference for the fur-ther research on the data analysis and forecasting in the random process.