Abstract: 【Purpose/significance】Exploring the importance of the characteristics of the academic innovation index system, providing
a basis for the study of academic innovation evaluation.【Method / process】Based on the systematic review of the research on academic
innovation evaluation, this paper extracts the characteristic indicators related to academic innovation based on the analysis of existing
achievements and the study on the connotation of academic innovation, and gives the calculation method of characteristic indicators.
After analysis, the dynamic network measurement method is selected to calculate the innovation power of the paper, and the correla⁃
tion between individual features and innovation power is analyzed through scatter plots to analyze the influence mechanism of individu⁃
al features on innovation power.【Result/conclusion】The characteristics of each index and the academic innovation of the paper are
not simply linear. It is not possible to simply use the traditional indicator system to evaluate academic innovation. It is necessary to ex⁃
plore machine learning methods for evaluation.【Innovation/limitation】Study the characteristic index system from a new perspective,
obtain the importance of each characteristic to the evaluation of innovation through empirical analysis of the machine learning model,
screen out the characteristics that are truly related to academic innovation, and correct the original index system.