Research On Macro-economic Forecasting Method Based On Multidimensional Data Fusion Of Electricity Consumption

Macro-economy is an important indicator to comprehensively measure the level of economic development of a country or a region.To some extent,it can also help China to better test the achievement of the first 100 year goal.Studying the development trend of macro-economy and making accurate short-term macro-economic forecast can help the government,enterprises and individuals to make correct economic decisions quickly.Macro-economic is complex.It is affected by many factors and power consumption is a very important factor among them.However,the prediction effect of the economic predic-tion model with only one factor considering is often unsatisfactory.Therefore,in the big data era,it is necessary to use advanced methods of big data and information technology,comprehensively consider various factors and integrate the macro-economic information hided in the data of these factors,so as to grasp the development trend of macro-economic more accurately and get more accurate macro-economic prediction results.Therefore,on the basis of the past macroeconomic research results and make the change of times into consideration,which can make the macro-economic forecast model conform to the development characteristics of the times,this paper selects electricity con-sumption as the basic factor,meteorology as the disturbance factor that influences electric-ity consumption and then affects macro-economy,online social network as the era factor to build macro-economic forecast model.This forecast model is a deep learning model of multidimensional data input,which is based on LSTM(long short-term memory)neu-ral network,optimized by PSO(particle swarm optimization)algorithm and aimming at studying the relationship between electricity consumption,meteorology,online social net-work and macro-economy.And then through the data of these three factors,the forecast model can get more accurate macroeconomic forecast results.In this paper,GDP is taken as the representative data of macro-economic,in order to analyzing the relationship be-tween these factors and GDP,electricity consumption of the whole society is taken as the representative data of electricity consumption,average temperature,average relative humidity and precipitation are taken as the representative data of meteorology,and the number of new microblog users is taken as the representative data of online social net-work.What's more,these data are also used to verify the PSO-LSTM deep learning neural network model.The final results show that there are bi-directional Granger causality and strong linear correlation between electricity consuption and GDP.There is a strong lin-ear correlation between the number of new microblog users and GDP and unidirectional Granger causality is existed from GDP to the number of new microblog users.And there are bi-directional Granger causality between these three types of meteorological data and GDP.What's more,on the basis of electricity consumption forecasting GDP,combining the data of average temperature,average relative humidity and new users of microblog can effectively improve the forecasting accuracy of GDP and obtain more accurate GDP forecast results,and this model can get better economic forecasting effect when it is used to predict the economy of provinces with more active online social activities.