Figure 2
From: Predicting the complexity and mortality of polytrauma patients with machine learning models
The construction of polytrauma mortality model, including its discovery, validation, performance evaluation and its comparison with existing scoring systems. The original dataset was randomly divided into a discovery and a validation cohort at a ratio of 7:3. In the discovery cohort, the SMOTE algorithm is used for sample balancing, and then models are built using SVM, RF, XGBoost and ANN models. The models are trained and tested using tenfold cross-validation to select the optimal model and perform feature importance analysis. Next, the predictive generalization and reliability of the model are validated in the validation cohort. The superior performance of the model was further validated by comparing its performance with the commonly used ISS, TI, and GCS scores in the validation cohort.
