Syndat: synthetic data quality evaluation
Syndat is a Python library to evaluate and visualize the quality of synthetic patient and clinical trial data. It computes synthetic data quality metrics for tabular datasets — Jensen-Shannon distance, normalized correlation difference and discriminator AUC — and aggregates them into three scores (distribution, correlation and discrimination) on a 0–100 scale. Visualizations cover feature distributions, correlation structures, summary statistics and SHAP-based discrimination plots, including stratified raincloud, GOF and trajectory plots for clinical subgroups.
Syndat is developed at Fraunhofer SCAI and used in the NFDI4Health consortium. The evaluation methodology is described in Adams et al., iScience 2025.
Start with the quickstart to install Syndat and run a first evaluation, or read the metrics page for an explanation of the quality metrics and how to interpret the resulting scores.
Contents: