An accurate ecological footprint analysis and prediction for Beijing based on SVM model

Accurate prediction of the ecological footprint (EF), an effective indicator for measuring urban sustainable development, enables better protection of urban ecosystems and alleviates discrepancies in urban development, resource utilization, and environmental protection. In contrast to previous research using general methods, we introduce the support vector machine (SVM) method. A novel model with improved prediction accuracy based on SVM is proposed, and we deploy this method to predict the EF of Beijing between 2016 and 2020. First, we calculate the EF of Beijing between 1996 and 2015 and screen out the 6 dominant indicators of EF changes using partial least squares (PLS). Second, based on 2014 and 2015 EF data, we compare the prediction accuracy of the back propagation neural network (BPNN) with the SVM using the 6 indicators as inputs and EF as the output, which then allows us to predict the year 2020 EF in Beijing. The results demonstrate that (1) the relative error rates between the prediction value and the actual value using the two models are 2% and 1% in 2014 and 3% and 0.53% in 2015, respectively, and the fact that the standard deviation of the SVM approaches zero demonstrates its higher prediction accuracy and stability compared to the BPNN; and (2) the EF of Beijing almost doubled to 8984 ten thousand acres from 1996 to 2015 and is predicted to increase to up to 14,206 ten thousand acres by 2020. Based on our prediction model, we provide science-based suggestions for the future development of Beijing.