时间序列组合预测模型在我国居民消费价格指数中的应用

发布时间：2018-05-26 07:50

本文选题：时间序列ARIMA模型 + 灰色GM(1　；参考：《兰州交通大学》2016年硕士论文

【摘要】：居民消费价格指数,英文名称consumer price index,即简称CPI,是普遍编制的一种指数,它可以用于分析市场价格的基本动态,为我国政府制定政策,对经济的宏观调控提供重要参考依据。为了能更精确地把握CPI的走势,提供分析依据,本文着重利用了一种组合模型—时间序列与灰色预测模型的组合,来对CPI进行预测。由于现代科技的不断进步与发展,各种预测方法随之与时俱进,因此发展了很多预测方法,就拿预测CPI的方法来说,就有很多,如时间序列模型、灰色模型、BP网络神经等方法层出不穷。但每个模型都有自己优势所在,当然也有它自己不可避免的缺点,所以为能缩小预测值与实际值得误差,使预测值的可信度更高些,本文将基于有效利用各种单一模型的优点,把不同模型的计算结果综合起来,根据误差大小分配单个模型在组合模型中所占的权重系数,相互取长补短,来弥补各种单个模型的缺点。本文选择两种模型进行组合,即在基于时间序列的组合模型分析方法的基础上,对我国的CPI进行建模预测。本文首先着重详细介绍了时间序列相关理论知识,紧接着运用这些理论对我国2013年5月—2015年4月CPI的月度数据对其进行建立模型,其次介绍灰色预测模型的建模理论,因为灰色预测模型需使用小样本数据,所以只选取了2014年7月—2015年4月的数据,建立模型并进行短期预测。建立单一的模型并通过检验后,然后,求得单个模型的绝对误差值,对2014年7月—2015年4月,分别求两个模型的绝对误差的平方和,利用方差倒数的方法算出两个模型在组合模型的分别所占的权重系数,也就是根据误差大小建立组合模型,误差平方和大的模型所占权值较小,相反,误差平方和较小者,其所占权值反而大。除了用数据说明组合模型的优势有降低预测误差的偏差大小以及波动幅度外,文中还用理论证实了此优势。最后对我国2005年—2014年的CPI数据建立基于时间序列的组合模型,并预测出2015年、2016年的CPI数据,预测结果表明,我国近两年的CPI较稳定,在政府制定政策时,进献上微薄之力。
[Abstract]:The consumer price index, the English name consumer price index, or CPI, is a general index. It can be used to analyze the basic dynamics of the market price. It provides an important reference for our government to formulate policies and to provide an important reference for the macroeconomic regulation and control. In order to more accurately grasp the trend of CPI, this paper provides an analysis basis. This article focuses on this paper. This article focuses on this paper. The combination of a combination model, time series and grey prediction model, is used to predict CPI. Because of the continuous progress and development of modern science and technology, various forecasting methods are progressing with the times, so many forecasting methods have been developed. There are many methods for predicting CPI, such as time series model, grey model, BP network God. Each model has its own advantages, but each model has its own advantages, and of course it has its own unavoidable shortcomings. So, in order to reduce the prediction value and the reality, the reliability of the prediction value is higher. This article will be based on the advantages of the effective use of a variety of models, the results of different models are integrated, according to the error. The weight coefficients of a single model in the combination model are complementary to each other to make up for the shortcomings of a variety of individual models. In this paper, two models are combined, that is, on the basis of the combination model analysis method based on time series, the modeling and prediction of CPI in China are carried out. The theory of sequence related theory is followed by using these theories to model the monthly data of CPI in China from May 2013 to April 2015. Secondly, the model theory of grey prediction model is introduced, because the grey prediction model needs small sample data, so only the data from July 2014 to April 2015 are selected and the model is established and carried out. A single model is established and the absolute error value of a single model is obtained by establishing a single model. The sum of the absolute error of the two models is calculated from July 2014 to April 2015. The weight coefficients of the two models in the combined model are calculated by the method of the reciprocal of variance, which is based on the size of the error. In combination, the weight of the square error and the large model is smaller. On the contrary, the weight of the small error square sum is larger. In addition to the advantage of the combination model with the data to reduce the deviation of the prediction error and the amplitude of the fluctuation, the advantage is confirmed by the theory. Finally, the CPI data of China from 2005 to 2014. A combination model based on time series is set up and the CPI data of 2015 and 2016 are predicted. The forecast results show that our country's CPI is more stable in the last two years.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：F224;F726

【参考文献】