The application discloses a
feature selection method,
system, device and medium for clinical prediction of
preeclampsia, relates to the technical field of
feature selection, and comprises the following steps: collecting clinical blood sample data of pregnant women in the middle of
pregnancy; extracting features in the clinical blood sample data, dividing search agent populations in the features into different clustering clusters, obtaining a guide vector from the
centroid of the clustering cluster and the global mean of the
population, performing aurora egg trajectory updating to obtain a current solution; fusing historical and current
population members to construct a
difference vector, applying disturbance to individuals, and obtaining an optimal solution under a set condition; converting a continuous
problem space of the obtained optimal solution into a binary domain, feeding a classification error rate back to a
fitness function, guiding a search process in a next generation
population, and outputting an optimal biochemical indicator combination with the highest performance and the best explainability until a termination process is satisfied, which indicates a future direction for the precise prevention and
personalized treatment of
preeclampsia.