Influencing factors analysis and forecasting of residential energy-related CO2 emissions utilizing optimized support vector machine

Abstract

With the economy development and rapid urbanization, the residential usage of energy has been increasing in China, leading to more CO2 emissions in resident sector. To predict the trend of residential energy-related CO2 emissions accurately, it is significant to analyze the influential factors. In this paper, 18 preliminary indicators are identified by grey relational analysis to prove their correlation with CO2 emissions firstly. To reduce the redundancy of data, 4 main components are extracted by principal component analysis as predicting input data of support vector machine (SVM). By adding chaotic mutation and nonlinear weight index, the improved chicken swarm optimization (ICSO) algorithm is proposed to optimize the parameters of SVM, hereafter referred as ICSO-SVM. Finally, the new hybrid model is applied to predict residential energy-related CO2 emissions in Shanghai, China. The simulation results in the forecasting accuracy demonstrate that the ICSO-SVM model outperforms the compared original chicken swarm optimization model (CSO-SVM), particle swarm optimization model (PSO-SVM), genetic algorithm optimization model (GA-SVM) and basic SVM. The rigorous influencing factors analysis and the outstanding performance in predicting CO2 emissions of ICSO-SVM model can offer relevant scholars and policy makers more breakthrough points of residential CO2 emissions abatement.