The Combination Forecasting Model Of Time Series In China's Consumer Price Index

The consumer price index, what is referred to as the CPI, is a commonly compiled index.It can be used to analyze the basic dynamic of the market price and provide an important reference for our government to make policy and economy macro-control. In order to more accurately grasp the trend of CPI and provide analytical basis, the paper focuses on the use of a combined model- a combination of time series and gray prediction model to forecast the CPI.Due to the continuous progress and development of modern science and technology, a variety of forecasting methods are developed.Take methods for forecasting CPI, there are many forecasting methods, such as time series model, gray model, BP network nerve and so on. But each model has its own advantages as well its own inevitable shortcomings. To narrow the error value of prediction and the actual value, and to make the credibility of the predicted value higher, this article will combine the calculated results of different models together based on the effective advantages of a single model, according to the size of the error to assign weight coefficient of a single model in the combined model and make up for the shortcomings of individual models. In this paper, we choose a combination of two models,namely on the basis of a combination of model-based analysis on time series, to make a model for forecasting the CPI.This paper introduced the theory and knowledge of time series, and applied them to found a model with the month data of China CPI, from May 2013 to April 2015. Then the paper introduced the founding model theory of the gray forecasting model and founded the model to have a short-term prediction with the data from July in 2014 to April in 2015,because the model required small sample data. After establishing each single model and the test, we respectively calculate the sum of squares of the absolute error of the two models and use the reciprocal variance method to calculate the weight coefficient of two models in combination model with the data from July in 2014 to April in 2015. Then found combined model according to the size of the error, the model which has small error sum of squares has a large share of the weight in the model, on the contrary, the one with smaller error sum of squares shares the bigger weight.In addition to illustrate the advantages of the combination model in reducing the size of the deviation and the volatility of the prediction error by the data, the article also confirms this advantage by theory.Finally, for the year data of CPI from 2005 to 2014, a combination model based on time series was established, and CPI data in the year of 2015 and 2016 are predicted. The prediction results showed that, China's CPI is stable in the latest two years and makecontribution on deciding policy for the government.