Understand your data.
Gretel Evaluate provides an out-of-the-box report on every model you train, so you can easily understand how your data performs on key metrics and compare results side by side.
- Synthetic Data Quality Scores (SQS)
Measure how closely the generated data maintains the statistical properties of the original dataset.
- Privacy Protection Levels (PPL)
Show the tunable privacy levels of each synthetic model.
- ML Quality Scores (MQS)
Assess the performance of synthetic data on classification and regression models.
- Gretel Benchmark
An easy-to-use low-code tool that benchmarks multiple datasets against multiple synthetic models.
Benefits for Developers
Eliminate evaluation bottlenecks. Accelerate model development with faster data feedback loops.
Improve data comprehension. Get a deep understanding of the technical metrics that matter most in your data.
Increase operational efficiency. Easily compare synthetic models and datasets throughout the ML lifecycle.
Verify data meets quality standards.
We do the heavy lifting. Gain rapid insight into your synthetic data performance with a high-level composite Synthetic Data Quality Score (SQS).
Review data quality. Ensure your synthetic data is as good—or better than—real-world data with statistical assurance verified with Principal Component Analysis (PCA).
No data field is overlooked. Correlation between every pair of fields is computed and compared between real-world training data and generated synthetic data.
Analyze data accuracy and performance.
At-a-glance reporting on data utility. See how your synthetic data compares to real-world data with a Machine Learning Quality Score (MQS).
Data analysis you can trust. Synthetic and real data are trained and evaluated on AutoML classification and regression models and the top models are compared.
Benchmark model performance.
Track experiments or monitor data drift by comparing multiple models and datasets with just 5 lines of code.
Get valuable measurements like Synthetic Data Quality Scores (SQS), dataset statistics, and training time all in one place.