Individual Prediction Explanations (XEMP)¶
Note
Time series projects do not support SHAP explanations.
XEMP-based explanations¶
XEMP-based explanations are a proprietary DataRobot method, available for all model types. They are univariate, letting you view the distribution of the effect each specific feature has on predictions. (SHAP, by contrast, is multivariate, measuring the effect of varying multiple features at once.) XEMP explanations are only available if SHAP is not supported by a model or experiment type; the appropriate Individual Prediction Explanation type is determined by DataRobot and made available when you select a model.
To access XEMP insights, click a model in the Leaderboard and choose Individual Prediction Explanation (XEMP) to expand the display. If prompted, click Compute Feature Impact.
After successful computation, the preview displays. See the DataRobot Classic documentation for full details on working with the preview, interpreting the display, and computing and downloading explanations.
