Make better predictions through actionable insights in Prediction Studio (8.3)
You can now detect deteriorating predictive performance of adaptive and predictive models through notifications in Prediction Studio. These notifications are actionable insights that guide you to models with low accuracy. You can use that guidance to discover predictive models that might have become stale and, therefore, require replacement, or to identify adaptive models with predictor issues.
The notifications provide status information about monitoring parameters such as low performance, uneven response distribution, and a lack of monitoring data for a model.

By acting on the notifications that appear whenever a problem occurs, you can limit the negative effect on your strategies and customer relations that models with low accuracy might have.
You conveniently access the notifications from the Prediction Studio header, in a model instance, or in the list of instances of a model rule, as shown in the following examples:


For more information, see Viewing Prediction Studio notifications and Prediction Studio notification types.
Previous topic Enrich your decision strategies with H2O and Amazon SageMaker predictive models (8.4) Next topic Sample, migrate, and test production data in a simulation environment (8.3)