Skip to main content
. 2021 Mar 2;10:586054. doi: 10.3389/fcimb.2020.586054

Figure 4.

Figure 4

The top 30 features importance were obtained by SHAP analysis in the first predictive model. A total of 45 features were used to classify 953 cases of COVID-19 coagulation dysfunction and 1,077 cases of normal COVID-19 coagulation. (A) The features importance of the first predictive model based on additivity (take the top 30). (B) The overall analysis of the influence of each feature (take the top 30).