Research On Carbon Intensity Based On Improved Extreme Learning Machine Algorithm

Nowadays,the issue of climate change is hot in the world.Especially the global warming caused by increasing carbon dioxide emissions is serious,which has threatened the survival and development of mankind.As the largest developing country,China is in a period of economy transition.But the promotion of post-industrialization will further aggravate the contradiction between economic growth and pollution emissions,making China face greater pressure to reduce emissions.In recent years,China has actively developed a low-carbon economy,vigorously advocated energy conservation and emission reduction,so the society has paid more attention to carbon emission reduction issues.How to effectively use energy,reduce carbon dioxide emissions,finish carbon emission reduction targets,and achieve a coordinated development of economy and environment is a key proposition for China to take the road of low-carbon economy and sustainable development.From the perspective of carbon emission efficiency,basing on the carbon dioxide emission data of China and its’ 30 provinces,this paper adopts stochastic frontier analysis(SFA)and translog production function to estimate and analyze the situation of carbon emission efficiency.Through nuclear density estimation and spatial distribution analysis,it is concluded that the carbon emission efficiency of all provinces in China has a certain rise and convergent trend.Moreover,there are large efficiency differences between the east,middle,west and northeast areas.Then,this paper comprehensively selects seven factors,including economic development level,industrial structure,energy consumption structure,urbanization level,foreign trade dependence and ownership structure,and uses SFA to quantitatively analyze the relationship between various factors and carbon emission efficiency.Based on the empirical analysis of carbon emission efficiency and its influencing factors,this paper starts with the essence of carbon emission efficiency SFA model and the definition of carbon intensity,combines the actual trends of carbon emission efficiency and carbon intensity in the three regions as well as their spatial distributions to find out the correlation between carbon intensity and carbon emission efficiency.Besides,SPSS is utilized to verify the impact factors of carbon emission efficiency also have an effect on carbon intensity.And the analysis shows that the carbon emission efficiency and carbon intensity in China and its’ provinces do have a relationship: the carbon emission efficiency and the carbon intensity have a reverse relationship in quantity and have the similarity in spatial distribution.Furthermore,the factors screening by carbon emission efficiency make a synclastic influence on carbon intensity.After determining the influencing factors of carbon intensity,an extreme learning machine algorithm based on improved particle swarm optimization(IPSO-ELM)is proposed in this paper,which uses the IPSO algorithm to optimize the input weight and hidden layer threshold of ELM.This hybrid model not only maximizes PSO’s global searching ability and ELM’s fast learning speed,but also overcomes the inherent instability of ELM.IPSO-ELM model is used to predict the carbon intensity China and its’ 9 typical provinces,and the results show that this model has a good fitting performance.The empirical results further demonstrate the proposed IPSO-ELM model and the method for factor selection are feasible and effective.Finally,aiming at the influencing factors,recommendations for national and regional carbon emission reductions are proposed accordingly.