Research On Combined Forecasting Model Based On Time Series

With the rapid development of China's economy,the continuous improvement of people's living standards,and the continuous acceleration of the internationalization process,our economic development is growing faster and faster,and the per capita GDP is also rising.Consumer price index,or CPI for short,is an important indicator used to analyze the basic trend of market prices in China,and also an important basis for national macro-control.According to the change of CPI,the state can make a series of adjustments of monetary policy and fiscal policy,so as to ensure the stability of the market and the well-being of residents.Therefore,this dissertation focuses on using the combination model to predict the CPI of Jiangxi Province.With the popularization of artificial intelligence and the continuous promotion of various algorithms,more and more prediction models have been greatly improved.For different research problems,there are also targeted models to predict.For example,for CPI prediction,there are ARIMA model,SVM model,BP neural network model,etc.Because CPI data is an index with many factors,this dissertation believes that a single prediction model can not fit the information contained in the series values well.In order to reduce the prediction error and improve the accuracy of the model,this dissertation puts forward the combination prediction model and the combination coefficient method,starting from the information characteristics of the sequence value itself,to find a suitable model,so as to better predict the CPI of Jiangxi Province.This paper first introduces the related knowledge of time series theory,laying a theoretical foundation for the following.Then,according to the linear and non-linear characteristics of CPI data in Jiangxi Province,puts forward the method of selecting the combination of linear model and non-linear model.According to the characteristics of the data,the classical least square method and MAE weight coefficient method in mathematics are used to combine,and the combined weight coefficient method is proposed,and then the combined model is established,so as to predict the CPI sequence values of Jiangxi Province from April 2019 to January 2020.Finally,according to the prediction results,the first mock exam and the combinedmodel's prediction effect are compared,and the prediction results of the combined models among different combination modes are found.Thus,an optimized model is found to predict the CPI of Jiangxi.The result shows that the first mock exam is better than the single model.The combined weight coefficient methods proposed in this paper is superior to the other weight coefficient method in the prediction error.