Understand the Impact of Each Predictor in Your Model

Predictors are field values that might influence your prediction. Are you overwhelmed by the number of predictors in your model scorecard? Unsure which ones to pay attention to? Now Einstein gives you an impact score for each predictor.

Who: This feature is available to admins with the Einstein Analytics Plus, or Einstein Predictions license.

Where: This change applies to Lightning Experience in Enterprise and Developer editions.

How: After a prediction is finished building, find it in your predictions list and select View Scorecard from the action menu.Predictions List View Scorecard

Impact is a number between 0 and 1 that represents the scaled weight or importance of a predictor.Prediction ScorecardImportance (or weight, depending on the model type) represents how significant a predictor is in the predictive model. Einstein builds predictive models using the model type that performs best for the data in the particular model. Since some models use importance and others use weight, it can be difficult to compare predictors across multiple models. That’s why Einstein gives you an impact score, which you can find on the scorecard Details page.

The scorecard also shows you the predictive model type that was used to build the model. For true/false fields, Einstein tests Random Forest and Logistic Regression model types. For number fields, Einstein tests Random Forest and Linear Regression model types.

You can view predictor impact visually on the scorecard Overview page. Here you can see the impact for the top 5 predictors in your model. To view impact for all predictors, click View All Predictors.

Top Predictors