本文为美国乔治亚州立大学（作者：Zhejin Hu）的硕士论文，共61页。

本文比较了线性约束下单变量和多变量时间序列对中国铜铝期货市场的建模和预测。对于单时间序列，在建立ARIMA模型之前，已经测试并执行了静态数据转换。对于多变量时间序列，在建立VECM模型之前进行了协整秩检验。在所选模型的基础上，预测结果表明，多变量时间序列分析比单变量时间序列分析具有更好的结果，说明利用多变量时间序列之间的关系可以提高时间序列预测的准确性。

This thesis presents a comparison formodeling and forecasting Chinese futures market of copper and aluminum withsingle time series and multivariate time series under linear restrictions. Forsingle time series, data transformation for stationary purpose has been testedand performed before ARIMA model was built. For multivariate time series,co-integration rank test has been performed and included before VECM model wasbuilt. Based on selected models, the forecasting shows multivariate time seriesanalysis has a better result than single time series, which indicates utilizingthe relationships among the series can improve the accuracy of time seriesforecasting.

1 引言

2 分析方法

3 结果与讨论

4 结论与未来研究展望

附录A SAS代码

附录B SAS输出

更多精彩文章请关注公众号：