Prediction of Acute Traumatic Coagulation Based on Interpretable Algorithm

This study aims to establish an interpretable machine learning model for predicting acute traumatic coagulation. We used the MIMIC IV database and extracted 2814 patients based on medical inclusion and exclusion criteria. Four machine learning models are established and results show that the ATC model based on XGBoost algorithm has the best performance, with an AUC of 0.95 and an accuracy of 96%. Then we used XGBoost model to calculate the contribution of each feature value to the model and Shap Value method to analyze the contribution of feature values to prediction from both the entire sample and a single sample.