Fig. 3

Model prediction performance by AUC and interpretation by SHAP method. Figure 3 shows the predictive performance of the model and its interpretation using the SHAP method. Figure 3-a is the ROC obtained using three machine learning algorithms and their corresponding AUC. Figure 3-b shows the variables sorted by the absolute value of the mean SHAP value. A high value means a high impact on the model output. Figure 3-c visualizes the different SHAP values ​​of each variable and their impact on the results. The gradient from red to blue indicates the ranking of the variable values ​​from high to low. The horizontal axis indicates the impact on the result. Figure 3-d further shows the application of the SHAP algorithm in a single sample. Different variables with different SHAP values ​​jointly affect the prediction results of the sample