Abstract: Abstract:The process of scientific research activities have a progressive, derived feature of complex structure.The scientific research innovation chain-data chain-publishing chain are closely connected and iterative.The evolution of data ecosystem and its model evolution is crucial to promoting scientific research innovation and expanding knowledge dissemination. On the basis of conceptual backtracking, this paper uses the theory of data ecosystem to elaborate the different data-driven logics of scientific research innovation and academic publishing. According to the different data thinking, data institution,data subject relationship, data management structure, data circulation environment and data management methods, the research data ecosystem can be divided into three gradual stages: closed, expanded and collaborative. Summarize the development trend of research data ecosystem from single to ecological, from fragmentation to system, from unilateral to collaborative, from point-to-point to integration, from manual to intelligent. A theoretical model for the evolution of scientific research data ecological model is proposed, and provide a new idea for promoting the overall quality improvement of each element and link of scientific research ecosystem with scientific data ecological governance.