Fig. 4

Comparison of AUC values of important features screened in three stages. a In the first phase, The SHAP method was used for important feature screening based on RF. white blood cell count, arthralgia, monocyte percentage, α1-acid glycoprotein, ferritin, and sore throat. The AUC value of the model was 0.9639. b In the second stage, the SHAP method, also based on RF, was used to screen out 6 important features: ferritin × platelet count, ferritin × lymphocyte count, ferritin × total protein, ferritin/urea, ferritin × erythrocyte sedimentation rate, and α1-acid glycoprotein/creatine kinase. The AUC value of the model was 0.9456. c In the third stage, on the basis of GBDT, the SHAP method was used to screen out the final 5 important features: arthralgia, ferritin × lymphocyte count, white blood cell count, ferritin × platelet count, α1-acid glycoprotein/creatine kinase. The AUC value of the model was 0.9916