Energy Consumption Forecasting Using Support Vector Machines for Beijing

Support vector machines (SVM) is a widely used method which can treat problems involving small sample, devilish learning, and high dimension. The current paper conduct a multivariate SVM in a total-factor production framework, and the GDP per capita, capital stock and labor are taken as the independent variables and the energy consumption is the dependent variable. The Gaussian radial basis function is taken as the kernel function, and then the energy consumptions of Beijing between the periods 1978-2008 are forecasted. The empirical results suggest that the multivariable SVM is valid in forecasting energy consumption.