An Improved Carbon Trading Behavioral Modelling Method Combining Discretized Statistical Analysis and Extreme Learning Machine

Carbon market, as an efficient emission reduction mechanism, has been widely implemented around the world. However, if not being designed properly, carbon market cannot fulfill its original intention of design under the influences of both internal and external disturbances. There is an urgent need for models or tools for ex-ante simulation analysis of the carbon market. Since in a market environment, disturbances will influence the carbon market dynamics through trading behaviors, it is important to accurately model carbon trading behaviors. In this paper, an improved carbon trading behavioral modelling method combining discretized statistical analysis and extreme learning machine (ELM) is proposed. Based on the trading samples of human participants, stochastic behavioral model can be constructed by using discretized statistical analysis method. However, the unevenly distributed number of samples make it difficult to obtain reliable statistical results in some discretized value intervals. Thus, based on the results of discretized statistical analysis, ELM is used to further improve modeling accuracy by utilizing its generalization capability. The case studies validate the feasibility of the proposed method.