[Purpose/significance] Scientific data fusion is a necessary means to activate the potential of data elements and boost the reuse of scientific data.At present, the research on scientific data fusion is relatively scattered, and there is a lack of systematic research on the influence mechanism of scientific data fusion.[Method/process] As an explorato? ry study, this article after the strict definition of scientific data fusion, by qualitative method of grounded theory, com? bined with depth interviews, based on scientific data life cycle theory, stakeholder theory and PEST analysis, construct? ing scientific data fusion SIE factor model,concrete cover nine main categories, category of 27 and 74 initial concept. [Result/conclusion] From the perspectives of stakeholders, internal data and external management, the research com? prehensively investigates the objects, processes, results and methods of scientific data fusion, and extracts five types of relations: promotion, foundation, decision, guarantee and intermediary.This framework makes up for the gap of dynamic mechanism and action mechanism in scientific data fusion research, and extends the research direction of model frame? work verification and fusion mode exploration.