Manage Prediction Definitions

Use the Model Manager to manage prediction definitions in your org.

• View a Prediction Definition
  Manage a prediction definition and its associated models.
• Configure Automatic Model Refresh for a Prediction Definition
  Use automatic refresh to schedule regular, unattended model updates on a weekly or monthly basis.
• Configure Automated Prediction Writeback for a Prediction Definition
  Automated prediction writeback saves numeric prediction scores to a field in a Salesforce object. Using this feature requires that you deploy a model that outputs a numerical prediction (regression) or score (binary classification) and for which you can map all model fields back to Salesforce fields. You can use an existing AI Prediction field or create a custom field by specifying a label.
• Configure Performance Monitoring for a Prediction Definition
  Configure performance monitoring if you want Einstein Discovery to display performance monitoring charts for a prediction definition. Performance monitoring is available for prediction definitions that use automated prediction writeback.
• Configure Model Quality Notifications
  Models drift when characteristics in the real world data diverge significantly from the training data used to build them. Einstein can send out alert notifications if your model performance falls below configured thresholds. Significant increases in prediction error rates, missing values, or out-of-bounds values can trigger notifications, as well as detected disparate impact among variables being analyzed for bias. You can set alert thresholds, schedule weekly or monthly performance evaluation cycles, and specify who will receive notifications.
• Configure the Model Card for a Prediction Definition
  Model cards help you document and communicate important usage information about your predictions. A model card shows statistics associated with the data that’s used to train the model. It can also show any optional explanations you provide about the prediction’s intended use, design assumptions, target audience, capabilities and limitations, and other relevant information. Disclosing these details helps users to understand predictions and differentiate among multiple predictions. Then they can make ethical, informed decisions about whether a prediction suits their use case.
• Analyze Prediction Accuracy with the Einstein Accuracy Analytics App
  The Einstein Accuracy Analytics app helps you monitor how well your models predict actual outcomes over time.