Research On The Prediction Of Consumer Price Index Based On Baidu Search Index

With the rapid development of mobile Internet and other technologies,Internet big data has gradually become a new strategic resource that enterprises and society pay attention to.At present,the methods and technologies of Internet big data for macroeconomic forecasting are receiving more and more attention from scholars and macroeconomic policymakers.Massive Internet search data not only represents the focus and degree of attention of netizens and consumers,but also can map social and economic development and reveal the behavioral trends of market players.Therefore,Internet search data can provide the necessary data basis for macroeconomic issues.The Consumer Price Index(CPI)is a macroeconomic indicator used to measure the degree of change in the price level of consumer goods and services paid by residents in a certain period of time.Important position.Therefore,CPI forecast research is of great significance.At present,CPI forecasting research focuses on the establishment of different forecasting models,mainly including regression models,time series models,machine learning models and combined forecasting models.Most of the data used in the CPI forecasting model are its own historical data and government statistics,and there are few researches on CPI forecasting using the method of searching for data on the Internet.Therefore,this paper introduces the Baidu search index to predict the CPI.This paper first introduces the related concepts of CPI and Baidu search index,and analyzes the relationship between Internet search data and consumer price index based on equilibrium price theory and consumer behavior theory.Secondly,combined with news text,long-tail keyword mining,demand map and other methods,select the network search keywords related to CPI to build a keyword database.After data preprocessing,the grey correlation degree between five initial keyword fingers and CPI is calculated.Then,use Spearman correlation analysis and time difference correlation analysis to screen out the first network search keywords with high correlation with CPI.On the basis of Baidu search index as explanatory variables,macroeconomic indicators are added,and the final keywords with predictive ability are screened out by stepwise regression as explanatory variables.Finally,in the empirical part of the CPI prediction model,the data is divided into two parts,the training set and the test set,and the regression model,the BP neural network model,the SVR model and the combined prediction model are constructed,and the relative error percentage,average absolute error,and average absolute percentage are introduced.Error and root mean square error observation standard deviation ratio are four evaluation indicators,and the test set data is used to verify and analyze the prediction effect of different models.The optimal forecasting model is selected through analysis,and the optimal forecasting model is used to forecast the CPI.The research conclusions of this paper show that:(1)it is determined that there is a strong correlation between the Baidu search index of online search keywords and my country's consumer price index;(2)based on Spearman correlation analysis and time difference correlation analysis,scientific and reasonable selection Predictive Power Keywords.After comparing and analyzing the evaluation indicators of the prediction models,it is found that among the single prediction models,the BP neural network model has the best prediction effect,followed by the SVR model,and finally the regression model.The combined forecasting model of the SVR model is the best among the four models,and the combined forecasting model RSR value is 0.2816,indicating that the forecasting effect is more accurate and the precision is higher,and this method can effectively predict CPI.(4)The Baidu search index has high timeliness.Using the Baidu search index to apply to the combined forecasting model can predict the CPI in advance,and can accurately predict the CPI change trend.