The invention relates to the technical field of clinical examination and diagnosis, in particular to a construction method and device of a traumatic
femoral head necrosis risk prediction model, and the method comprises the steps: determining candidate diagnosis markers; screening out differentially expressed candidate features as a preliminary screening
feature set based on the candidate diagnostic markers; constructing a traumatic
femoral head necrosis risk prediction model by using a plurality of differentially expressed candidate features in the preliminary screening
feature set and a preset
machine learning
algorithm to determine key biomarkers, forming a final
feature set, and updating the traumatic
femoral head necrosis risk prediction model; therefore, the problems that in existing research, due to
neglect of the influence of
nonlinear correlation possibly existing between diagnostic markers and the synergistic effect in
disease development, a generalized and universal prediction and diagnosis model cannot be established, and then deep exploration of a potential
molecular biology mechanism and accurate evaluation and intervention of the early stage of diseases are limited are solved.