Research On Urban Unemployment Problem Based On Combination Forecast Model Of ARIMA-Double Weight Neural Networks

On March 5,2018,the first meeting of the thirteen National People's Congress was held in Beijing.At the meeting,Premier Li Keqiang of the state council of the People,in his government work report,reviewed of the past five years work mentioned:more than 66 million new jobs in cities and towns,the population of more than 1.3 billion countries to realize relatively full employment;In the formulation of the main target for the development of 2018,it is pointed out that more than 11 million new urban jobs will be created,the urban unemployment rate will be less than 5.5%,and the urban registered unemployment rate will be less than 4.5%.Prom these data of the government work report,we can see that our attention on urban unemployment problems of China.Not only that,many scholars are studying the urban unemployment related problems.And this article is through the adoption of a new prediction method to study the urban registered unemployment number and urban registered unemployment rate in our country.In this paper,we firstly use the double weight neural networks model and ARIMA model to predict the urban registered unemployment number and the urban registered unemployment rate respectively.Then,the two models are combined into a combina-tion prediction model according to the optimal weights,using this model to forecast and analyze the urban registered unemployment number and urban registered unemploy-ment rate in our country.The results show that:for the urban registered unemploy-ment number in our country,when the weight of double weight neural networks model is 0.5287,the weight of ARIMA model is 0.4713,we can get a combined model while had better fitting effect,the result of prediction are 9.80747 million people in 2017,9.858164 million people in 2018,9.875616 million people in 2019,and 9.880789 million people in 2020;for the urban registered unemployment rate in our country,when the weight of double weight neural networks model is 0.613,the weight of ARIMA model is 0.387,we can get the combined model fitting effect is best,and the predicted results are 4.1054%in 2017,3.9705%in 2018,4.1150%in 2019,and 3.8898%in 2020.Accord-ing to t,he analysis,it,is concluded t,hat the fitting effect of the combination prediction model is not only better than the double weight neural networks and ARIMA model fitting effect,but also the BP neural network.which also shows that the new method used in this paper can better predict the urban registration unemployment problems of China.