Figure 3
From: Predicting the complexity and mortality of polytrauma patients with machine learning models
The performance metrics of the mortality prediction models. (A) The mean value and the standard deviation of the performance indicators, including Accuracy, Recall, F-score and AUC, of four ML models (SVM, RF, XGBoost and ANN) using ten-fold cross-validation was shown as the bar chart. (B) The performance indicators of the optimal ML model for predicting the mortality risk in the validation cohort were compared against the commonly used severity scores, including ISS, TI and GCS. (C) The top 15 features that contribute to the polytrauma mortality model classification were shown. HGB, Hemoglobin; TRF, Thoracic rib fracture; SH, Superficial hematoma; RBC, Red blood cell count; TVH, Thoracic visceral hematoma; BAS%, Basophil percentage; HR, Heart rate; HCT, Hematocrit; MCHC, Mean corpuscular hemoglobin concentration; BAS#, Basophil count; MPV, Mean platelet volume; Tca, Total calcium; MONO%, Monocyte percentage.
