Model Bias Mitigation

Perform the following steps to score a model from the View Model Catalog page:

1. In the View Model Catalog page, click Action next to corresponding model and select Bias Mitigation.
   The Model Bias Mitigation drawer appears.

   Figure 11-64 Model Bias Mitigation

   
   

   

   
   
2. Select the fairness metric from the Select Fairness Metric drop-down menu.
3. Select a scoring metric from the Select Scoring Metrics drop-down list.
4. Select a protected feature from the Select Protected Features drop-down list.
5. Select a conda environment from the Conda Environment drop-down list.

   The selected Conda environment is sourced from the model context and is passed through to the backend for each operation so that the appropriate environment can be activated before executing the Python code. The Conda environment can be selected for models that are built from model catalog or model pipeline.

6. Select the average type from the Average Type drop-down list.
7. Click Execute.