BMC Medical Informatics and Decision Making

Table 5 Utility metric results

From: Comparative assessment of synthetic time series generation approaches in healthcare: leveraging patient metadata for accurate data synthesis

Generation Method	Model^a	TSTR metric
A1	DGAN_T	0.802 ± 0.033
	WGAN-GP_T	0.880 ± 0.002
	DGAN_M	0.931 ± 0.064
	WGAN-GP_M	0.909 ± 0.065
	DGAN_P	0.983 ± 0.005
	WGAN-GP_P	0.986 ± 0.002
A2	DGAN_T	0.900 ± 0.589
	WGAN-GP_T	0.837 ± 0.109
	DGAN_M	0.869 ± 0.078
	WGAN-GP_M	0.924 ± 0.049
	DGAN_P	0.906 ± 0.083
	WGAN-GP_P	0.919 ± 0.000
A3	DGAN_T	0.919 ± 0.167
	WGAN-GP_T	0.900 ± 0.156
	DGAN_M	0.893 ± 0.079
	WGAN-GP_M	0.758 ± 0.125
	DGAN_P	0.957 ± 0.016
	WGAN-GP_P	0.778 ± 0.120

Values in bold highlight the best-performing approach per trained model and metric
TSTR Train on Synthetic and Test on Real
^aThe subscripts on the model column refer to the dataset used for the training step

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com