The Application Of Web Search Data On Predicting Consumer Confidence Index (CCI)

The rapid development of the Internet has made network big data a new strategic resource for enterprises and society.In the era of big data,the use of big data methods and technologies for macroeconomic forecasting is receiving more and more attention from scholars and macroeconomic decision makers.A variety of network data hides consumers' various behavioral habits.These data can map out economic development status and consumer psychological expectations.Extracting the required information from relevant technologies is a hot topic in current research.This paper,through the quantification of network search data,predicts the consumer confidence index(CCI)to reflect the consumer confidence situation.The paper analyzes the key influencing factors,provides guidance for the preparation of consumer confidence index and related economic issues,and has practical significance.The content of this paper is mainly divided into three parts.The first part,through introducing Consumer Confidence,Consumer Information Needs and Behaviors,the paper constructs a theoretical framework of the relationship between Consumer Confidence and Web-Search Behavior.It also introduces the choice of network data source and analyzes the advantages of network data source in forecasting.The second part,this paper combines the compilation principle of the National Bureau of Statistics Consumer Confidence Index and related literature to set initial keywords and thesaurus,and introduces keyword screening methods.In the data processing of keywords,this paper analyzes the consumer confidence index data and Baidu search data from January 2011 to October 2018 nationwide,and initially screens 181 web search keywords through relevant literature and uses them.The time difference correlation analysis method and the LASSO algorithm perform dimensionality reduction to obtain core search keywords for prediction.In the third part,the Kernel Partial Least Squares Regression model,Decision Tree Regression model and Random Forest Regression model are constructed and compared.In the analysis of each prediction model,the comprehensive prediction effect of the Random Forest Prediction Model is better than the other two models by comparing the prediction effects within the model sample and outside the sample.The Random Forest Prediction Model helps to improve the accuracy.Moreover,the degree of keyword importance of Random Forest output also contributes to the weight decision making of traditional Consumer Confidence.This paper adopts the more popular machine learning prediction algorithm in the era of big data,which enriches the application of network data in economic forecasting.It also improves the forecast system of Consumer Confidence Index and makes up for the lag of traditional data.It analyzes the influencing factors of consumer confidence and proposes revisions to the weighting problem in the compilation of traditional Consumer Confidence Index,which helps to improve the accuracy of compilation.