An evaluation of time series summary statistics as features for clinical prediction tasks

Table 23 Performance of 1-year mortality prediction by multiple classifiers

Combination	AUROC	AUPRC	AUROC	AUPRC
	Logistic regression		Random forest
mean	0.7704 ±0.0025	0.7398 ±0.0030	0.7838 ±0.0017	0.7324 ±0.0034
first	0.7505 ±0.0043	0.7171 ±0.0024	0.7561 ±0.0018	0.7031 ±0.0028
min, max	0.7759 ±0.0036	0.7404 ±0.0046	0.7844 ±0.0011	0.7298 ±0.0020
min, max, mean	0.7776 ±0.0033	0.7413 ±0.0015	0.7840 ±0.0022	0.7330 ±0.0049
min, max, mean, std	0.7899 ±0.0024	0.7534 ±0.0040	0.7872 ±0.0020	0.7391 ±0.0030
max, mean, std, Q1, kurt	0.7998 ±0.0007	0.7642 ±0.0028	0.7876 ±0.0014	0.7408 ±0.0026
	SVM		Decision tree
mean	0.7996 ±0.0015	0.7634 ±0.0018	0.6367 ±0.0019	0.5653 ±0.0037
first	0.7694 ±0.0024	0.7343 ±0.0026	0.6218 ±0.0006	0.5495 ±0.0024
min, max	0.7923 ±0.0024	0.7552 ±0.0024	0.6368 ±0.0026	0.5654 ±0.0013
min, max, mean	0.7959 ±0.0023	0.7607 ±0.0021	0.6375 ±0.0030	0.5658 ±0.0014
min, max, mean, std	0.7978 ±0.0019	0.7619 ±0.0017	0.6409 ±0.0028	0.5689 ±0.0023
max, mean, std, Q1, kurt	0.8090 ±0.0013	0.7734 ±0.0007	0.6370 ±0.0018	0.5654 ±0.0028

ISSN: 1472-6947