中国消费者价格指数应用研究——基于调和分整自回归移动平均模型  

Application Research of Chinese Consumer Price Index Based on ARTFIMA Model

作　　者：王锦 商豪[1] WANG Jin;SHANG Hao(School of Sciences,Hubei Univ.of Tech.,Wuhan 430068,China)

机构地区：[1]湖北工业大学理学院,湖北武汉430068

出　　处：《湖北工业大学学报》2020年第1期106-109,共4页Journal of Hubei University of Technology

基　　金：湖北省教育厅人文社会科学研究项目(14G191).

摘　　要：针对中国居民消费者价格指数时间序列数据呈现出的记忆特征,采用描述性统计分析和经典时间序列分析相结合的方法,建立两种不同的长记忆模型,分析比较两种模型的谱密度函数拟合图和三期预测值误差,结果显示ARFIMA模型的平均预测绝对误差为37.34%,而ARTFIMA模型的平均预测绝对误差为28.57%,因此ARTFIMA模型的拟合预测效果更好。由此得出中国居民消费者价格指数序列更加符合半长期记忆性的结论。The Autoregressive Tempered Fractionally Integrated Moving Average(ARTFIMA)time series model is applied in this paper.Aiming at the memory characteristics of the Chinese Consumer Price Index data,a combination of descriptive statistical analysis and classical time series analysis is used to establish two different long memory models,and the spectral density functions of the two models are compared.The combined graph and predicted value error show that the ARTFIMA model has better fitting effect and prediction effect.This concludes that the Chinese consumer price index sequence is more consistent with the semi-long-term correlation.

关 键 词：时间序列 长记忆性 ARTFIMA模型 ARFIMA模型 

分 类 号：O212[理学—概率论与数理统计]