Downstream Utility Evaluation

Classification

We provide comprehensive tools to evaluate whether your synthetic data maintains utility for real-world applications, particularly in machine learning workflows. When your models achieve comparable performance using synthetic data versus original data, it demonstrates that the synthetic data preserves the essential patterns and relationships needed for downstream tasks.

Logistic Regression

See usage in rockfish.actions.EvaluateLogisticRegression.

Random Forest

See usage in rockfish.actions.EvaluateRandomForest.

Time Series Forecasting

We also provide tools to evaluate the utility of synthetic data for time series forecasting tasks. In the context of synthetic data evaluation, forecasting measures how well models trained on synthetic data can predict future values, reflecting the data’s usefulness for real-world predictive applications.

See usage in rockfish.actions.EvaluateForecast.