As with any deployed predictive analytics tool, model governance and monitoring model performance is critical to reduce risk and maintain effectiveness. Predict has several ways of following these protocols depending on your specific needs.
Predict Flows are configured with guardrails in place to make sure that results stay within desired performance bands. If models do experience significant drift, they can be retrained to accommodate the changes in the input data.
Models deployed in Predict can also be evaluated against a baseline model, or a simplified model, to ensure that their performance is providing a substantial benefit.
Upload Data