Factor analysis and forecasting of CO2 emissions in Hebei, using extreme learning machine based on particle swarm optimization

In the prevailing low-carbon economy, China is under enormous pressure to control CO2 emissions, therefore, of great significance is the study to analyze what influential factors mainly contribute to emissions, so as to forecast emissions accurately and harness the growth from the source. In this paper, basing on 22 influencing factors identified by bivariate correlation analysis, factor analysis is then adopted to extract the latent factors which essentially affect emissions and 8 special factors transformed by scoring coefficients are acquired. Extreme learning machine (ELM) whose input weights and bias threshold were optimized by particle swarm optimization (PSO), hereafter referred as PSO-ELM, is established to predict CO2 emissions and testify the availability of the factor analysis. Case studies reveal that the factor analysis which generates 8 factors as input can highly improve prediction accuracy. And the simulation results demonstrate that the built model PSO-ELM outperforms the compared ELM and back propagation neural network in forecasting CO2 emissions. Eventually, the analysis made in this study can provide valuable policy implications for Hebei's CO2 emissions reduction and strategic low-carbon development.