Prediction of carbon flux based on wavelet networks

Low-carbon, the only way to the sustainable development of all countries around the world, has become a hot topic. Carbon flux (FC) is closely related to many factors in ecological environment as an index of global carbon emissions. Therefore, it is very important to find effective methods to study the relationship between FC and environmental factors. A predicted model based on wavelet networks is proposed in this paper. And it is compared with BP neural network and support vector machine (SVM) on network structure, predicable accuracy, convergence rate, and so on. The experimental results show that wavelet network is more stable and accurate, and can get higher convergence rate.