An evolutionary cost-sensitive support vector machine for carbon price trend forecasting

This paper aims at the imbalanced characteristics and proposes a novel evolutionary cost-sensitive support vector machine (CSSVM) by integrating cost-sensitive learning, support vector machine, and genetic algorithm for carbon price trend prediction. First, carbon price trend prediction is converted into a binary-class prediction problem for CSSVM, in which a higher misclassification cost is imposed on the minority samples. In comparison, a more negligible misclassification cost is imposed on most samples. Second, a genetic algorithm (GA) is used to optimize all parameters of CSSVM synchronously. Taking Beijing, Hubei, and Guangdong carbon markets as samples, the empirical results show that the proposed model has a higher classification accuracy and lower misclassification costs compared with other popular prediction models. Furthermore, the sensitivity analysis verifies that the proposed approach is robust.