Research On Construction And Application Of PMI Attention Index Based On Network Information

The Purchasing Managers Index,or PMI,is a forward-looking monthly index used to monitor the state of economic development.It plays an important role in macroeconomic forecasting and business analysis.Although PMI is more advanced than other government departments’ statistical reports,it still has shortcomings: obtaining data in the form of questionnaires,the cost of data acquisition is too high;the data are qualitative data,which are greatly affected by the subjective emotions of the respondents;the sample size and Changes in sample composition will affect index trends and so on.With the development of the Internet,a large amount of quantitative data can be extracted from network information.This article takes the manufacturing PMI as an example,and uses the network information to replace the questionnaire to build a manufacturing PMI attention index to make up for the lack of manufacturing PMI.The index reflects the feasibility and application value of macroeconomic changes.This article sorts out the internationally accepted manufacturing PMI concepts and compilation methods,and compares and analyzes the survey methods and compilation contents with China’s manufacturing PMIs in line with China’s national conditions,and points out the deficiencies in the compilation of Chinese manufacturing PMIs.In order to overcome these shortcomings,this article constructs the PMI attention index of the manufacturing industry.First,the word segmentation algorithm is used to process the text information of China HowNet CSSCI Periodical Database and Guotai’an News Database to construct a keyword thesaurus.Then the crawler algorithm is used to obtain the keywords.Baidu Index,which uses Baidu Index to calculate keyword search heat to remove the influence of network development trends on data samples;combines Kmeans clustering and factor analysis to calculate the attention diffusion index;finally,extracts important feature scores as random attention model weights through the random forest model and Synthetic Manufacturing PMI Attention Index.In order to test the effect of the manufacturing PMI attention index,a long-term relationship analysis was performed using the data of the manufacturing PMI attention index for the 20 th,25th,and 30 th consecutive days and the manufacturing PMI data;at the same time,the EEMD-BP forecasting model was constructed.The 20-day average of the manufacturing PMI attention index and the three-phase lagging data of the manufacturing PMI are used as input information to predict the manufacturing PMI for the current month.The prediction effect of the manufacturing PMI attention index is verified by the data from January 2011 to September 2019.The research results show that: first,the keywords of network information corresponding to the manufacturing PMI can provide predictive information highly related to the questionnaire;second,the synthetic manufacturing PMI attention index has a longterm stable relationship with the manufacturing PMI,indicating that manufacturing Industry PMI attention index can describe the development trend of the manufacturing industry to a certain extent;third,adding the manufacturing PMI attention index of the current month to the model can further improve the prediction accuracy,indicating that the manufacturing PMI attention index contains the foreseeable information in the future.The accumulated 20-day information of the PMI attention index compiled based on network information is relatively stable,which can be 10 days earlier than the official release of manufacturing PMI information,which is more time-effective.