Abstract: In the context of accelerating the construction of a unified national data element market in China, the design of a data transaction and circulation model that suits the country’s specific conditions, along with a corresponding revenue sharing mechanism, has not yet been fully explored. This paper starts by setting up a data element transaction model based on the commonly recognized three main parties in the data element market, using meteorological data supply as a typical subdivided industry and the application of data in power forecasting as a typical scenario. The model incorporates a wind power prediction model based on machine learning, depicting the value realization of multi-feature data elements and data services. Furthermore, in designing a revenue sharing mechanism based on data value, this study compares the differences in the main benefits to data producers (data vendors) and market impact among four revenue sharing methods: the average method, leave-one-out method, Shapley Value method, and Penalty-modified Shapley Value method. Lastly, the research demonstrates that the Penalty-modified Shapley Value revenue sharing strategy effectively considers the level of differentiation among data elements in the market, while also identifying and preventing disturbances caused by data element replication.

Key words: data element circulation, data factor market, revenue sharing, machine learning, Shapley value