Skip to content

Testing Your Model

TruEra's Test Harness is designed to ensure a standardized approach to model evaluation. The framework lets you specify expected model behavior for comparison to actual results with respect to current and future data collections.

Using either TruEra's Python SDK or the TruEra Web App, you can explicitly create model testing definitions for:

  1. Acceptable performance targets for models based on different data
  2. Model performance on different segments of data
  3. Fairness criteria satisfied by models on given data
  4. Expected stability of model predictions on various data
  5. Acceptable number of unimportant features to a model on given data.

See Model Test Summaries for guidance on interpreting the test results.

Click Next below to continue.