MARMA（p, q）模型理论研究及应用.pdfVIP

华东师范大学硕士学位论文 摘 要 当家装公司制定第二年预算时，仅用销售额时间序列或开单量时间序列这种 单变量时间序列已无法满足需求，很多时候需要综合考虑多个因素的互相影响， 因此矩阵型截面数据的时间序列应运而生。本文提出的矩阵型截面数据时间序列 MARMA(p, q)模型在使用范围上较前人所提的矩阵型截面数据时间序列 AR(p) 更广泛，预测精度也更准。 本文首先将矩阵型截面数据时间序列的相关概念进行了详细描述，并提出了 矩阵型截面数据时间序列MA(q)模型。 然后，本文又提出了矩阵型截面数据时间序列 MARMA(p, q)模型，并给出 其定义，与多维的时间序列ARMA(p, q)模型的等价性条件，平稳性条件，模型 识别及参数估计等方面进行了探讨。 最后，本文通过家装行业 2 个门店的销售变化率数据用多维的时间序列 ARMA(p, q)模型与矩阵型截面数据时间序列MARMA(p, q)模型进行模型构建， 发现用 MARMA(p, q)模型可以使模型的残差平方和变得更小，更能很好地刻画 家装行业多对象多属性之间的关系。 关键词：矩阵型截面数据时间序列，MARMA 模型，参数估计。 i 华东师范大学硕士学位论文 Abstract When home decoration companies make the second year budget, only using single object time series like sales time series or order time series is unable to meet the demand. A lot of times people should consider the effect of many factors, then matrix cross- section data time series have emerged. MARMA(p, q) model of matrix cross-section data time series which this paper given has the wider range than AR(p) model of matrix cross-section data time series the previous presented, and the accuracy of forecast is more precise. First, this paper given the detail description of matrix cross-section data time series concept, and presented MA(q) model of matrix cross-section data time series. Then, this paper also presented MARMA(p, q) model of matrix cross-section data time series, and given its definition, the equivalence condition with the multivariate ARMA(p, q) model, the stationary and reversible conditions, model recognition and parameter estimation. Last but not least, this paper constructed multivariate ARMA(p, q) model and the MARMA(p, q) model of matrix cross-section data time series through the sales data rate of one decoration company, It can be found that using the MARMA(p, q) model of matrix cross-section data time