Carbon dioxide (CO 2) is the main cause of the greenhouse effect. With the rapid development of the economy in China, CO 2 emissions have increased dramatically. To reduce CO 2 emissions, ensure the sustainability of China's economy and implement the Paris International Convention, it is important to investigate the main factors affecting CO 2 emissions and use those factors to accurately forecast CO 2 emissions. In order to achieve accurate prediction of CO 2 , this paper proposes a CO 2 emission prediction model based on principal component analysis (PCA) and particle swarm optimization for least squares support vector machine (PSO-LSSVM). Through data 1990-2016 in Hebei Province of China, this paper identifies 24 influencing factors though the bivariate correlation analysis. After applying PCA to reduce the dimensions of the influencing factors, two principal components were extracted as input variables. Then the parameters of the LSSVM model are obtained by PSO and the forecast model is established. By comparing the prediction results with actual values, it is proved that the prediction error of the PSO-LSSVM prediction model is 0.663%, which is smaller than that of the traditional BPNN and LSSVM models.