Machine learning techniques for mortality prediction in critical traumatic patients: anatomic and physiologic variables from the RETRAUCI study

Table 5 Performance properties of the 9 algorithms analysed

Algorithm	Accuracy	Specificity	Precision	Recall	F-measure	AUC
LR	0.901	0.970	0.685	0.459	0.623	0.912
CT	0.899	0.966	0.647	0.438	0.603	0.856
JRip	0.899	0.962	0.638	0.462	0.624	0.730
BN	0.894	0.955	0.603	0.469	0.630	0.915
NN	0.889	0.952	0.576	0.451	0.612	0.890
SMO	0.901	0.978	0.710	0.366	0.533	0.672
ADABOOST	0.892	0.971	0.630	0.337	0.500	0.891
BAGGING	0.902	0.968	0.668	0.444	0.609	0.910
RFOREST	0.901	0.964	0.650	0.460	0.623	0.905

LR Logistic regression model, CT Classification tree, JRip Repeated Incremental Pruning to Produce Error Reduction, BN Bayesian network, NN neural network, SMO Sequential Minimal Optimization, ADABOOST Adaptive boosting, BAGGING Bootstrap aggregating, RFOREST Random forest, AUC Area under ROC curve. In bold values with statistically significant differences

ISSN: 1471-2288