Research On Improving The Accuracy Of Economic Forecasting With Internet Search Data

In the study of economic forecasting,it has always been a difficult problem to the Outliers’prediction.However,the emergence of outliers often indicates that there is a major mutation in the economic trajectory.Thus,outliers prediction is an important issue for the research of economic forecasting.For conventional time series model,outliers caused by mutation events can not be accurately forecast.Because of the lag of economic indicators publication,when the occurrence of mutation events or before it,factor models or the time series models included factors can not predict the outliers effectively.It lead to an enormous loss in the economic activities,or missing an important opportunity.In recent years,the research of the Internet information technology which improve the predictive ability is developing rapidly.In this paper,based on the previous research,combining the cases of China’s economic development,we make an exploration to improve the effectiveness of outliers prediction by the Internet information technology,which is based on the traditional model.The key points are the outliers’ direction caused by mutations and the accurately of the outliers prediction.In this paper,we make empirical research on the China’s construction machinery market.At first,We gathered the relevant historical data,and then set up the time series model based on the data analysis.Model established in this paper is a seasonal difference auto-regressive moving average(SARIMA)model.On this basis,we try to introduce the Internet search data into SARIMA model.In this paper,we defined the basic Internet research indicators from three economic entities based on market demand theory in economics:producers,consumers governments.Then,according to the semantics,we expanded the benchmark indicators and collected the data.Since the Internet research indicators’ data set is large,in order to simplify the model and improve the usefulness of the model,we use the statistical methods to reduce indicators’dimensions.Then using principal component analysis to reduce the dimension and introduce the main components to the SARIMA models.Finally,we compared the forecasting effect of the SARIMA model with the SARIMA model which contain Internet searches.And the result showed that,compared with the traditional model,the model evolved Internet search data has a better prediction to the mutation,and the directionless and accuracy in both research objectives achieved,the predicted overall error rate was also significantly decline.