Research On The Prediction Method Of Urban Registered Unemployment Rate Based On Different Data Sources

Unemployment is one of the most common phenomena in modern ecnomic society.The actual situation of people’s unemployment is mainly reflected by the unemployment rate.The state pays particular attention to the urban unemployment situation.Therefore,it is necessary to conduct research on the urban unemployment rate in order to assist the government branches to complete the relevant policies which would promote employment.At present,in our country the most widely used index to measure the rate is registered unemployment rate in towns.In fact,nevertheless,this index has some issues,for instance,long publication time period,insufficient statistical depth,and multiple data distortions.Therefore,this article is based on different data sources for urban registration.Unemployment rate forecasting methods are studied.Based on different data sources from both macro and micro perspectives,review of China statistics data and big data are selected to predict the rate.The former is based on the establishment of a model which is based on the correlation between the rate of unemployment and the GDP development.The yearly rate of the unployment and economic development from 2010 to 2019 in the statistical yearbook are selected to study the GDP development and the unemployment in the east,middle and west.The following three models are established based on the hierarchical Bayesian method:Fay-Herriot model,time series model,and generalized Fay-Herriot model based on time series.The three models are simulated by Gibbs algorithm.Together,the optimal time series model is obtained from the convergence and stability of the model.The latter uses the monthly and annual Baidu search data and urban registered unemployment rate from2012 to 2019 to establish the lagged variable model(ADL),the multivariate mixed data autoregressive model(MADL-MIDAS)and the mixed frequency convolutional neural network model(SMCNN)),the results show that according to the internal and external errors of the sample,the SMCNN model is better than the other two models considering the forecast effect.Comparing the of perfect search and accurate search,the SMCNN is superior to others.So the SMCNN model can be more accurate and comprehensive get the inflection point of the unemployment rate.Through a comprehensive comparison of the unemployment rate based on different data sources,the following conclusions are drawn: first,data from different data sources has its own advantages.Big data is social,open,and real-time,so a model is built based on this data source.The obtained forecast data is more real-time.Second,based on the statistical yearbook data to study the urban registered unemployment rate,the unemployment situation can be researched subregional.According to the current review of China statistics,data of various provinces and cities can be obtained so that unemployment can be studied by region.Third,building SMCNN models through big data can comprehensively and accurately obtain inflection points.The turning point is the development trend of the unemployment rate and its structural change point.A exhaustive and precise capture of the turning point can better seize the development trend of the unemployment rate.Based on the above research results,when the government obtains the data of unemployment,it can fully mine the information that be obtained from different data sources,that is,not only obtain data through offline statistical surveys,but also can through online data.Study the unemployment rate so as to realize the guiding role of unemployment rate data for government work.