Fig. 3

The pre-selected feature sets of the optimal model were evaluated through SHAP. a Features are listed in descending order according to contributions for the XGBoost-2 in predicting suicide attempts. b The feature effects on identifying suicide attempts. The color indicates the values of the features from high to low. The horizontal location shows whether the effect of the value leads to the prediction of suicide attempts. Each point is a SHAP value for a case and a feature. c The decision plot of XGBoost-2 predicting suicide attempts. Each line represents a case. From the bottom of the plot to the top, SHAP values for each feature are added to the base value of model, and each line strikes the x-axis at its corresponding observation’s predicted value to obtain prediction results. *IGT: IOWA gambling task, SST: suicide stroop task, HAMD-24:Hamilton Depression Scale-24 items, CTQ: Childhood Trauma Questionnaire