Research On Carbon Emission Prediction Model Of Construction Industry Based On Machine Learning

Nowadays,the global carbon emission problem is becoming more and more serious.China is actively promoting the low-carbon and green development of various industries.The construction industry is the key area to achieve the goal of energy conservation and emission reduction.The basic work to realize the low-carbon development of the construction industry is to measure and predict the CO₂ emission of the industry.Improving the accuracy of the CO₂ emission prediction model is of great significance to help the construction industry to achieve energy conservation and emission reduction.In this paper,based on BP neural network and support vector machine algorithm with nonlinear processing advantages,the intelligent prediction model of carbon dioxide emission in construction industry based on machine learning is studied.Firstly,this paper analyzes the current situation of carbon dioxide emission in the construction industry,defines the boundary of carbon dioxide emission in the construction industry,establishes the calculation model of carbon dioxide emission in the construction industry,obtains the data of carbon dioxide emission in the construction industry in each year,selects 12 Factors of carbon dioxide emission in the construction industry through the literature research,and carries out the influence factors based on the random forest algorithm It is found that the completed area of buildings in construction industry is the most important factor,and GDP,total output value of construction industry,completed area of buildings in construction industry,labor productivity of construction industry,number of employees of construction enterprises,primary energy consumption of construction industry are determined as input variables of carbon dioxide emission prediction model of construction industry.Secondly,aiming at the deficiency of traditional prediction model in carbon dioxide emission prediction of construction industry,this paper puts forward two prediction models BP neural network and support vector machine and simulates them,and compares and analyzes the prediction results of three models including traditional ARIMA model with four indexes of average relative error,average absolute error,root mean square error and².The results show that the four error indexes of the traditional prediction model are larger than the other two models,and the prediction results deviate from the actual values seriously,which indicates that the machine learning prediction model is superior to the traditional prediction model;among them,the average relative error of the prediction value of the support vector machine model is the smallest,and the prediction results are the closest to the actual values.Finally,in order to further improve the prediction accuracy,the chaos particle swarm optimization?CPSO?and fuzzy cuckoo search algorithm?FCS?are applied to the BP neural network and support vector machine prediction models respectively to optimize the relevant parameters in view of the current BP neural network weight and threshold selection problems and the lack of support vector machine parameter selection methods.The simulation results show that the prediction accuracy of the optimized model is higher than that of the unoptimized model,the average relative error of the optimized SVM model is reduced by 2.68%,the average relative error of the BP neural network prediction model is reduced by 9.75%,the accuracy of the CPSO to the BP neural network is improved more,but the accuracy of the optimized SVM prediction model is improved It is still higher than BP neural network.Then,the FCS-SVM prediction model with the highest accuracy is selected to predict the carbon dioxide emissions of China's construction industry in 2019-2023.The results show that the total carbon emissions will rise in the next five years,but the growth rate will slow down significantly,and the carbon dioxide emission intensity of the construction industry will continue to decline before2023,but the decline rate is very low.